Supply Chain Management Bootcamp

1. Class Introduction : So whether you are working in an e-commerce company or in an offline retail business, having a good understanding of supply chain can help you build and grow your business. And that's why I've made this class on the introduction to supply chain management. So if you're new to this class, hi, my name is now beep. I am an MBA graduate and the co-founder of blood. In the past, I've worked with multiple e-commerce and health care companies. I've helped them build their own supply chain and operation. And this class, we will understand about the different components of supply chain, like procurement, logistics, manufacturing, and demand planning. Now to make sure that you use these concepts in the real-world, I'll be showing you the supply chain processes of companies like Tesla, Amazon, and Apple. Now we will start this class by building a strong foundation. And we will understand about the basics of supply chain and the different types of supply chain we have. So we will understand about the closed loop supply chain, the linear supply chain, pull and push kind of supply chain, etc. Then we will understand about bullwhip effect, risks and challenges in supply chain. And there's lot more to cover. So whether you are a working professional or a college graduate who wanted to explore their career in supply chain, then this is the right class for you. If you're excited for this class, Let's dive in. 2. What is Supply Chain ?: So hey, everyone. Now we are starting our first section of the course, and the main purpose of this section is to help you build a strong foundation about supply chain. So I'm going to cover a lot of concept in this section, but the main goal is to just build a strong foundation. Now, whenever I start any section in any of the course, I usually try to oversimplify the concept by giving you super simple example. In this course, I'm going to cover the five main topics that are important in supply chain and then I'm going to explain those concepts like I'm explaining it to a 5-year-old. First, in this section, we'll understand about supply chain and operation management and what's the difference between both of them. Then we'll understand about supply chain metric. Then we'll understand about the different types of supply chain and how do you decide the trade off between fast, flexible, and cheap. In the end, we'll understand the different components or functions of supply chain. Let's start with our first concept where I'm going to oversimplify all the concept, I'm explaining it to a 5-year-old and the main purpose is that I wanted to build a strong foundation for everyone and that's why this is just a oversimplified explanation. Let's start with a simple story and I love explaining a concept with a story. This story is about how Apple goes from a farm to your table. So imagine you open a grocery app and you order some apples. Now, if you carefully think about it, behind the scene, there are many things that are going on pari. For example, there is a farmer in some corner of the world who is growing those apples for you. Then he is loading those apple into the truck. The truck is then picking those apples and moving them to the nearby warehouse, and then they are moving it to the nearby retail store. And that's how your apple goes from your far to your front door or on your table. Now the thing we are most curious about is how exactly all of these things work together so well. Now to make it happen, you need to understand about operation management. Operation management is like running a retail store or running a kitchen or running a dance studio. It's like, how do you operate a specific system in the supply chain? It could be a warehouse, it could be a retail store. So imagine you run a grocery kitchen and you decided, Hey, what kind of ingredients should I keep into the stock? How fast should I deliver? And if I have 34 people, who exactly does what and when? So you need to answer all of these question. So in operation management, we will make sure that everything runs smoothly in our system with no mixer or hiccups and delay. So when you go inside a kitchen, the chef will make sure that the food quality is great. The kitchen manager will make sure or ensure that chef always have all the ingredients, and the cleaning staff ensures that it is always clean so that people doesn't have to waste their time cleaning things up by themselves. So the main idea of operation management is how time, task, and team work together to deliver faster and better result. And this could be applicable for a retail store, for a warehouse, for a kitchen. But you might have one question in mind. How do you measure if things are working properly in your supply chain or in operation management? And that's where your supply chain metric, they come into the picture. A couple of metric that are super important and are very obvious in supply chain is on time delivery. If a company is promising you 24 hour delivery, are they delivering all of their orders on time or not? Then you have inventory. Are you storing too much inventory in your warehouse or you are having a sokout situation because you are storing too less then cost per delivery and a bunch of other metric. Now, in simple terms, these are also known as KPIs or key performance indicator. In this course, as we go along, we will be creating a lot of KPIs dashboard and I'll be explaining you how do you solve each of these metric? Now, the main purpose is that this metric will help decision making simpler for the company. Now in this section, we'll also understand who does what exactly in supply chain. Like when you look at a grocery store, then they have a person who actually procures the item, then there is one person who plans which product to put in which shelf, and then there's a person who manages the day to day operation. There's a person who just accept payment, and there's a person who delivers the grocery. And you need to make sure that each role fits like a puzzle to make sure that the experience is amazing. So great. Enough introduction about all the concept in the first video now we'll go step by step and learn about each and every concept in oversimplified manner because I feel that people come from different background and different understanding about supply chain, and I want it to make it super beginner friendly as well. 3. What Is a Supply Chain?: So let's start with our first concept. What exactly is a supply chain? And I'm going to stick with the same Apple example that we were discussing in our first video. So let's discuss about the day in the life of an apple. So let's say you open a grocery app and you ordered some apple, you waited for almost 20 minutes and our rider knocks on your door and boom. You got some apples at your doorstep. But if you think carefully behind this small incident goes a long list of process that has to work together. For example, there is one farmer who is growing apples for you, a wholesaler is buying these apples from all these farmers and storing in the warehouse in good condition, then they are moving these apples to a fulfillment center where these retail stores are buying apple in large quantity, and finally, a person is picking, packing, and delivering these orders for you. And all of this is possible because of supply chain. When you look carefully at a supply chain, it's a team spot. Consider supply chain like a relay race where the supplier hands the bat into the transporter, the transporter for their heads, the bat into the warehouse, the warehouse passes it to the last mile delivery driver or a retail store. And finally, you receive the product. Now at each step of the process, they need to ensure that the handoff is quick, smooth, and efficient. That means they want to minimize the delay any friction, and they want to ensure that the things are happening on time without wasting much money or time. If any one of these components slows down, it affects the whole supply chain. Let me help you understand this with the help of a very simple example. Now, we experience supply chain in our day to day life without actually realizing it. For example, go to your washroom or bathroom and look at the toothpaste that you are using. I'm sure that the ingredients that are there in your toothpaste came from at least five to ten different countries. Even a T shirt that you're wearing right now have gotten who is grown in India, died in Vietnam, and might have stitched in Bangladesh. In fact, the device in which you are watching this video right now, it might not have even manufactured in your own country. So everything is superconnected around the world with the help of supply chain. And the main question we need to understand in this course is how do these companies make sure that the whole chain runs smoothly every single day? And to understand that in the next video, we'll understand about operation management. And operation management will help us understand how each of these supply chain process runs smoothly. 4. Operations vs. Supply Chain – What’s the Difference: So let's understand what is operation management and how is it different from supply chain? And let's go back to our same example that we were using in our last video. What happens behind the scene in your local grocery store? So consider supply chain team like a delivery crew. They make sure that all the items in your retail store like apple, bread, milk, it arrives on time in the morning at your store. But the operation team works like a kitchen crew. They decide how much to stock, how to manage people, and the processes, and how to run the retail store efficiently. So consider supply chain like a back end, operation team works on the front end. And both of these team works together to keep the store alive and create a good experience for the end consumer. An operation team runs the store like a clockwork so as soon as the goods arrived in the retail store or in the warehouse, the operation team takes over. They make sure that the shelf are stocked neatly. They are managing the staff schedule. They are keeping the store clean, and they're making sure that the shelf are not empty or they are not overflowing with a lot of items. The operation team ensures that the experience of yours in the retail store is smooth, fast, and pleasant. Let me give you an example so that you understand how your supply chain team and operation team works together. Let's say our supply chain team delivered 200, crates of strawberries on any particular day. But the store doesn't really have space to store all of them. Or operation team didn't plan the capacity of these many strawberries in the retail store. And this is a big chaos. And that is why supply chain team and operation team needs to work together to make sure what arrives in a retail store, at what point in time. And then operation team handles that item and decides the player place where it needs to go. So that's the difference between operation management and supply chain. In the next video, we'll understand, how do we know if the operation team or supply chain team is doing their job properly or not? To understand that, we need to talk about supply chain KPIs or key performance indicator. But before that, I wanted to cover one topic on local versus global supply chain. 5. Global vs. local supply chains: So in this video, we'll understand about global supply chain and local supply chain. Let me try to explain this concept with a super simple example, and I love giving example so that I explain a topic like a story. The primary purpose is to oversimplify all the concepts for you. Let's say you walked into a nearby grocery store and you picked some juicy tomato. Now think for a second and try to understand how will you differentiate if these tomato came from your country or a nearby farmer or from a different country. If it came from a local farmer, then it's a part of a local supply chain. But if it came from a different country like Spain, Mexico, China, or any other country, then this tomato is a part of a global supply chain. Now, both local and global supply chain brings tomato at your retail store, but with a different root, time, and risk. And we'll understand more about these three important factors. So when we talk about local supply chain, because it has a short distance, it gives you all the items that requires freshness and they need it quickly. Also, if there is a problem with the supply chain of these items, you can easily track and resolve the issues in case of a local supply chain. So when you look at products like milk, bread, grain, most of these items come from a local supply chain, including the tomato. And you can get it at a very cheaper price because you have no import duty whenever you import item from a different country. So all the fresh items or perishable good usually follow a local supply chain. But on the flip side, you also have a global supply chain of some exotic fruits like Kiwi some variety of coffee beans or tuna. Many countries ship their best product to a different country at a premium and that's how your global supply chain operates. Now, things do break in global supply chain because of logistic issue, transportation issue, tariff, and multiple factor. Global supply chain has a risk, but if people want more variety of product or a different taste. In that case, they have to rely on global supply chain. It gives you more variety and volume, but local supply chain makes it more fast and fresh. So local supply chain means freshness and speed, and global supply chain means variety and scale. But you might have one c in mind. How do you know if your supply chain is working perfectly fine? Like, how do you measure the health of your supply chain? Like, are you getting product delivered to your warehouse or retail store on time or not? Or are you selling enough item in your retail store before they expire? To understand all those concepts, we'll talk about key performance indicator, also known as KPIs, which will simply track your retail store, score the individual metric so that you can improve your supply chain and your store metric. 6. Supply Chain Goals – Cost, Speed, Service: So great. Now in this video, we'll understand about the three main goal of a supply chain cost, speed, and service. So imagine you run and manage your own grocery retail store. Your primary goal as a manager and owner of a retail store is to make sure that your shelf are full of fresh item. You have happy customer who wanted to come back to your retail store and you're making good profits without wasting money. Now, to make all these three things to happen, you need to make sure that your costs are low, your delivery is efficient and quick, and you are offering great service to your customer. Now the problem or the bottleneck is that you cannot maintain all three at once. Because if you are maintaining a really low cost, you cannot provide a great service. If you are providing a great service, then you cannot maintain a low cost. So when you look at your retail store, you have all of these costs in your retail store. Maybe let me take a laser pointer so that I can explain all of these concepts. The first cost is procurement cost. You need one person in your retail store that procures individual item on your behalf from different companies at a cheaper price. So they procure from companies at a wholesale price. That's your procurement cost. You need one or two people in the admin staff to make sure that they procure all the items. Then you have inventory cost. When you store any item in your retail store, until that item is not sold, well your capital is reserved, that's your inventory cost, and you're not earning any interest on this capital. Third one is production cost. This is only when you produce something by your own. Let's say you're having a small team of people who actually cut and dice fruits and that kind of comes in production. Then you have warehousing cost. If you have four or five retail store and you maintain a warehouse, then you need to store this item in a warehouse. You need to pay rent, electricity, people. Then you have transportation cost and distribution cost if you're delivering the product to the doorstep. So things like transportation costs, storage cost, spolage or labor or handling, all of this cost will slowly add up and it will eat your margin. Now, there are different ways by which you can reduce down the cost, like you can buy in bulk at a lower price. You can optimize your delivery at a specific time, or you can save on fuel or store less. But all of these things also bring a good amount of risk. Now your second goal when you operate a retail store or a warehouse or anything is speed. How speedly can you restock your item or deliver your product? It depends on the goal of that specific operation. For example, if you maintain a warehouse, then speed means how quickly can you refill the stock. When you maintain a retail shop, then speed also means how quickly can you deliver the product. Speed matters a lot in retail because imagine you're running a retail store and suddenly you are out of milk, bread, butter, all these essential items in the early morning. I mean, people will just come to your retail store and they'll not buy anything because you don't have essential items that are needed for that purchase. So good supply chain will help you refill your shelf on the same day. They respond quickly whenever there is a demand or a spike, um, and they usually don't want customer to walk away because you don't have anything. But the problem is, anytime you need a faster delivery from your partner or from your company, they doesn't offer you a good price. So you need to plan properly for your retail store or for your warehouse. The third thing is service. And when I talk about service, it means your shelves are no were empty, you have fresh product, you delivers on time and your store is clean and you have well managed stuff. Now the biggest problem or the tradeoff is that you cannot win all three. You can win two at a time. For example, you have quick, low cost and high quality. Either you can have low cost and speed, low cost and high speed, or you can have low quality and low cost, but not high speed. In the upcoming video, we'll understand our real world supply chain function from buying a good to delivering them to the end user. 7. Functions in Supply Chain management: So hey, everyone. Now we are starting a new section in the course, and in this section, we'll discuss about functions in supply chain management. Now in this section, we are going to cover three important concepts. First, I'm going to explain the importance of planning and coordination in supply chain. Then we will talk about reverse logistic. Whenever you ship a product to end user or to your warehouse, there is always a chance that that product might come back because either it is damaged or it's not in a proper condition, we'll understand about reverse logistic returns and repair. In the end, I'm going to explain all the key stakeholders that play an important role in a supply chain. So stakeholders like vendors, three PL, and distributors. Now, let's start with our first concept to explain this concept in a easy way, I'm going to stick with the same example that we were discussing in the last few videos. We were talking about how exactly the supply chain of apple that you purchase online or from a retail store works. Let's say a customer picked a juicy apple from a nearby grocery store, and this sound very simple to most of us. But if you look at the exact process that Apple goes through in the whole supply chain, it includes purchase, operation, logistic, resource management, and information. It goes through a lot of things in the supply chain. You can categorize these processes into four major steps. The first one is procurement, where a retail chain or a distributor is simply procuring the raw material, or in this case, apple from a farmer or a supplier. The second step is manufacturing. Now sometimes manufacturing means that you are manufacturing a new item from raw material, but many times it also means that you are simply washing it or cutting the fruits into smaller pieces so that you can just simply sell. Or you can make a different product like Apple jam or apple drink out of the product. Then you have logistic where you are transporting the product from a warehouse to a retail store or from a distributor to the warehouse. And in the end, you have fulfillment, where the product is actually reaching to the customer doorstep with the help of a delivery partner or a customer is walking into your retail store. These are the four backbone of supply chain functions that you normally see in a retail supply chain. So these are the four backbone function of a retail supply chain, and let's discuss about each of these function one by one. Let's start with procurement and manufacturing. Procurement simply means that you are selecting the best supplier. For example, you might have a list of five to ten different supplier who actually supplies you different variety of apples from different parts of the country or the world. So you need to pick some local supplier, some international supplier who supplies you the product. And that price negotiation also requires you to decide how much quantity are you ordering and how quickly do you need it? And then you are placing the order. So that's your procurement process where you go through all of the suppliers that you have sort listed and you finalize the quantity and speed of delivery. So that's your procurement. Then comes manufacturing. And manufacturing could be simple, just like you're washing the apple or it could be complex where you are cutting the apple, packing it into some boxes, or you're making a different product like fruit jam or something else from these apples. In the end, you have logistic and fulfillment. Logistic is when you are getting the product from point A to point B. So either you have multiple distributors or warehouse where you actually procure these apple and then you distribute worldwide or in your own country in different retail location. And fulfillment is when you receive the stock, you check the freshness of the stock, then you move the stock into the shelf, and then you label it and just sell it to the end consumer. A simple example of logistic and fulfillment is when you are receiving a batch of apple Great, and you're simply logging it, checking the quality, and then just selling it to the end user. So if I summarize the video in a supply chain, you have multiple function that works together in a perfect sync. Things like procurement, manufacturing, logistics and fulfillment. Now you might have one question in mind. Who ensures the reliability of this supply chain where everybody is supercnected, and they execute their plan on time? To understand that in the next video, we're going to talk about the importance of planning and coordination in the supply chain. 8. Role of Planning and Coordination: So now we'll understand about the importance of planning and coordination and supply chain. So even if you have the best suppliers, the fasest truck, and the cleanest store, but without a proper planning, your supply chain will fail. And that's why planning is super important. Let me help you understand this with the help of a very simple example. So imagine you suddenly received 200 crates of banana and your store only have a space of 50. And not just this. Let's say, in your store, you started promoting mangoes and you're running offers, but somebody in the planning team or the procurement team forgot to place the order of mangoes. And that's where planning is important. It literally connects everything. So when it comes to planning, what exactly are we talking about? So there are four major types of planning in a retail supply chain. You have demand planning, where you are simply predicting how much your customer will buy next week. You have supply planning, where you are simply estimating how much should we order from our supplier. Then you have inventory planning, where you simply try to understand how much stock do we keep at each location in our retail store and in the warehouse so that we can efficiently manage all the operation in time. In the end, you have delivery planning where you simply try to predict how much time it will take for us to ship a batch of a certain product from our centralized warehouse to a retail store. So these are the four important types of planning that usually happen in a retail supply chain. Well, planning is great. If you're able to predict how much your customer will need in the future, how much should you order, how much should you store, and how much time it will take for you to deliver the items from your warehouse to the retail store, it's great. But planning is not just about numbers. You also need to have a good coordination, and it's all about alignment. Like, you need to know about your supplier, how much they can produce and what's their turnaround time. Let's say you want your supplier to give you 20,000 water bottle in just one week, but their capacity of manufacturing is just 1,000 water bottle a day. In that case, you will have a higher lead time. So your buyer needs to talk to the supplier to understand how much they can produce, what's the quality, and what is the lead time. Also, whenever you place order to your manufacturer or supplier, you also need to inform the warehouse team so that they can store a certain good and they can plan to ship that specific item or product to the nearby retail location. And not just that, your retail store also need to prepare how much demand they can expect in a specific season. Like in every festival, there are certain set of products that are sold multifold. So you need to plan for their demand. And apart from that, logistic team also need to plan their routes, their timings, and how they're going to ship all of these product to different retail location. Now if these four teams do not coordinate well with each other, you can expect some delay, shelf might go empty, or you might end up overstocking a certain item and you won't be able to run your operation and supply chain efficiently. Planning and coordination will make the supply chain more efficient and seamless. Okay. So now you have a really good understanding about why coordination and planning is super important in supply chain. But there is one small problem. What if you ship something to your own retail store or to the end user, and that specific item comes back to you because either it was damaged or it was not in a good condition. So in the next video, we'll talk about what do we do if item comes back to you and how do you handle that thing efficiently? We'll talk about reverse logistic, returns and repair. 9. Reverse Logistics, Returns, and Repairs: So perfect. Now in this video, let's talk about reverse logistic, returns and repair. And how exactly does a retail store handle all of these complex thing in their supply chain? I I start from basic, in every single retail store, everything moves in this direction. A retail store buy things from supplier, then they sell it to the customer. And most of the time, customer buys it, keep it, and just use it. But sometimes things come back. And there could be multiple reason. Now, whenever a retail store or ecommerce warehouse get a return, they simply put that product back in the supply chain. It's like the backflow of the supply chain and the product will go back to the supplier or sometime even to the manufacturer. But what happens when a product is returned? Let's understand about the process a bit. So when something comes back, the store team will simply inspect the product first to understand if they are the one who actually sold the product to the end consumer because I've seen customer doing some fraud where they actually returns a product that they never purchased from that retail store. Then they decide that, hey, do we want to restock the item? Do we want to do a refund to the user, or are we going to send this product to the vendor or our supplier, or let's just discard it because it's not worth sending it back. Then they simply discard it or disposit then they record it in their system so that they can justify that damage or that return to their vendor or supplier. This way, they are also updating their inventory or filing a claim or simply tracking refund. So let's say a customer bought three carton of milk and the customer wanted to come back and return the item. So after inspection, if you realize that the product is still in a good condition, then you can simply refund the customer and you can restock it back. If you feel that the product inside those carton was slightly damaged, or it was a bit unsafe to consume, then you can just send it back to the vendor. And if it was spoiled, in that case, you are left with no option but to simply discard and then just disport the cartons. Who is handling the reverse logistic in a retail store? If the product is going out, it's the responsibility of the fulfillment team. If the product is coming back, then there's always a one to two people team in a retail store that handles the return and the damage and the disposal of item. But you might have one question in mind. Who are the key partners who help us manage procurement, delivery, return, and storage? Let's understand about all the stakeholder in the supply chain. We're talking about vendors, three PL, and distributor. In the next video, we'll talk about all of these key stakeholder in the supply chain. 10. Key Stakeholders – Vendors, 3PLs, Distributors: Now we'll talk about all the key stakeholder in a supply chain, and we're talking about vendors, three PL, and distributor. I'll help you understand what exactly this three PL is. And to make it easy for you to understand, let me give you a super simple example of the same retail store we were discussing in the last few videos. So when you walk into a retail store nearby and you look at it, everything looks effortless. But behind the scene, you have so many partners that work together like an orchestra to make sure everything works smoothly. So let me introduce the three most important stakeholder in a supply chain. Vendor, distributor and TPL, or third party logistic. Apart from Walmart, when you look at a small retail chain like Shopight. I was in the New York almost three months back and I saw a retail chain that was quite popular. The name of the retail chain was ShopRight and it was a small retail chain in New York and it has multiple location. So when I started learning about that retail chain, I realized that it was not massive retail chain like Walmart, Costco, Best Buy. It was a smaller one. And they work with a lot of distributor, third party logistic companies to make sure that their supply chain is efficient, cost effective, and they don't really have to put billions of dollar into every small part of the supply chain. So let's understand a bit about vendor, third party distributor, sorry, vendor, distributor, and third party logistic provider. Let's start with vendor. Vendor are all those people who actually supply some product. These could be farmers, a brand that is actually manufacturing item or some local bakery who is delivering fresh bread. Anybody could be a vendor that is actually providing you item from which you are procuring your raw material. Now, the procurement team will actually reach out to these vendor, negotiate the prices, manage the quality, plan the schedule, handle all the complaints and shortages. So that's the responsibility of a procurement team. And they try to maintain a really good relationship with vendor, and they always work with two, three vendor in case if somebody goes out of business or if somebody is not able to fulfill the demand on time. So to minimize risk, your procurement team works with multiple vendors. Then you have distributor and third party logistic network. These are simply your movers and middlemen in your supply chain network. Distributor will simply buy from multiple vendor, store in their warehouse, and then ship it to your on regional warehouse or directly to retail store. Their main purpose is to make sure that you don't really have to handle the complexity in the supply chain. Like there are distributor that store and distribute product of just one single brand. And you can directly buy from them at an affordable price without actually storing the item. Let me give you a small example so that you understand why do you need a distributor? For example, let's say you are a retail store in a city and you have ten different location and there is a regional distributor of Nestle. Now, this regional distributor might buy every single product produced by Nestle in millions of quantity, but you just have ten retail location. You can't really sell millions of quantity. So you'll say, Okay, I need 50,000 quantity of this product every single month and I'll try my best to sell it from these retail location. So when you buy 50,000 quantity, you may not be able to get the best price, but when they buy millions of quantity, they are getting a much cheaper price. But because they are buying millions of quantity, they are taking a risk. They are storing the inventory, and they are distributing that inventory to people like you. And that's why scale is important. If you are a small retail chain, you cannot afford to buy a large quantity, and you always have to work with the distributor so that you can buy at an affordable price and maybe add your own margin. And when you work with a distributor, you also need to work with the logistic provider. For example, for these ten location, let's say you have to work with maybe 60, 70 distributor and you cannot really own and operate your own fleet and truck. You might work with a company that actually provide all of these items from the distributor warehouse to your own warehouse and they work with ten of your kind retail store or retail chain. In that case, they doesn't cost you much, and that's why you need to use a third party logistic provider. Because you are not Walmart and you cannot afford to have your own and you cannot afford to have your own storage, warehouse, and logistics. These three party has to work in sync to make sure that you're getting all the stock on time, you're getting the fresh product, you're storing them better and selling it to the end user. If any of these stakeholder in your supply chain doesn't fulfill their promise, your supply chain breaks. There is a proper score model which can help you manage your supply chain better. I'll spend some time on this concept a little later in the course. 11. Overview of Metrics & KPIs: So hey, everyone. In this video, now we will understand how do you measure supply chain performance? The main reason of measuring the performance is because if you cannot measure something, then you cannot improve. Let's understand this with a very simple example of running a grocery store. So let's say you run a grocery store, but you do not measure how much inventory do you have? What's your daily sales? And because you do not measure your metric, you always either end up overstocking something or many things goes out of stock. That's why measuring the retail performance, the supply chain, inventory, procurement, demand, all of this is super important. Now, your supply chain uses KPIs or key performance indicator to answer some of the most common question. So imagine you're running a retail store. In every single retail store, you are expecting some deliveries from your supplier. So you'll always measure things like, am I getting deliveries on time in my retail store that I anticipated? Are all my shelves full that mostly a customer need? Are we holding too much inventory or too less inventory? And if I'm delivering something to the customer doorstep, how much does that delivery cost? So you have to measure your key performance indicator supply chain uses KPIs or key performance indicator to benchmark if you're doing the right thing or not. These metric are just like fitness tracker for your supply chain. They will show how healthy is your supply chain and what needs to be fixed if something is broken. Now I'm going to talk about a bunch of KPIs that you need to major in your business, whether you have a retail store, a warehouse, a distributor, or a manufacturer. All of these KPIs and metric are super important. The first one is OTIF. OTIF stands for on time in full, it simply answers you a very basic question. Was the delivery on time and was it completed, and did we got everything that we ordered? On time in full. Whatever we ordered, did we go all the quantity on time or not. For example, let's say you ordered 50 crates of milk and it arrived by 9:00 A.M. In the morning. If it doesn't come by 9:00 A.M. Or if it doesn't have 50 crates, then it's not on time in full. So it's a simple yes and no. So usually, when you look at any retail store or warehouse, they always try to benchmark OTIF around 95%. Now, this is not a standard baseline, but companies use OTIF metric to understand if their supplier or distributor is delivering the promised inventory at a given time or not. That's the main purpose of OTIF. Second one is inventory turn. Let me make it very clear. I'm not going to explain you all of these metric in super detailed fashion. We have proper assignment and exercise where I'm going to explain you how do you use these KPIs and metric to actually solve a problem and improve your business over time? I'm just giving you a high level explanation right now in this video. The second one is inventory turn. Are we selling our inventory in our retail store or in our warehouse? Is the inventory moving fast enough? Because anytime you hold some inventory in your retail store, you're bearing that cost. Inventory turn will simply help you understand how many times are you selling and replacing your inventory stock in one year. And this is on product level. This specific metric will simply help you answer, are the items in your retail store or warehouse moving fast enough or they are gathering dust? Do you have too much inventory sitting idle and it's not moving? Remember, if you have too much of inventory, you're holding too much of cash. If you have too less of it, it will frequently go out of stock, which is a bad thing. Let's say you have bananas which are just sitting in your retail store, nobody's buying and after a week, they will spoil and you just have to throw them away. So you have to make sure that your banana inventory turn should get over into three days, otherwise, you have to throw them away. But on the flip side, let's say you're selling vinegar. Vinegar can last for a quarter or even for a year. In that case, your inventory turn for vinegar might be 30 days or 45 days, but you cannot afford to do that with banana. So you have to figure out what's the lifespan of a certain item and what could be the most efficient inventory ton. In fact, in case of retail store, they actually sell 90% of whatever fruits and vegetables that they order on the same day and they just replenish that inventory on the next day except the exotic fruit. That's your inventory turn. If you have high inventory return, you have efficient cash flow. If you have low inventory return, you're wasting the space and your money is also at risk. Then you have CPD and fill rate. CPD is your cost per delivery. So you need to first calculate the wholesale price of your product, then you need to calculate the retail price and then you need to add delivery. And if you combine everything, you need to understand if you are actually selling or delivering a product at a profit or not. CPD is simply your cost per delivery. What is it taking you to deliver a product to the customer doorstep? It could be for a ecommerce platform, for a retail shop, it could be for anything. And Filter rate will simply help you understand what percentage of customer demand is being met. So if ten customer came to your shop, they demanded ten different product, how many products are there in your retail store? So maybe 80% or 85% fill rate. That means you have 85% of the product that customer demanded in your retail store or on your warehouse. I think that's a good fill rate. So these two metric are also important. As a business, your primary goal is that you should have a high fill rate and low cost per delivery. And this is very obvious. So perfect. You at least understand the basic foundation of these metric. I know it's difficult to put these things into in K study or in assignment, but don't worry, I'm going to give you enough K study and assignment in the course. But for now, these are the four important metric that can simply help you understand at least the basic concept about a retail store or in a supply chain. 12. Introduction to Cycle Time, Lead Time, and Service Level: So here, everyone. Now we're going to discuss about three important metrics that can make our process of order to delivery really efficient and fast. So from the time you place an order to the time you fill yourself, these three important metric will make your overall supply chain more efficient. And these are cycle time, lead time, and service level. And let me explain these things with a very simple example. Imagine you're ordering fresh strawberries for your grocery store. Now, generally, all fresh fruits and vegetables are restocked in every single retail location every single day early in the morning. But imagine item like chocolate or spices or any other product. They usually get replenished maybe in one week, two week or sometimes in a month. To understand time taken and the metric from the time you place the order to the time you replenish it in yourself. The first metric is cycle time. Cycle time shows you how long does individual process take. For example, how much time does it take to unload the truck and moving it to the shelf, how long does it take to pick a product from the shelf and pack it for the customer? That's your cycle time. Then you have lead time, which is how much time does it take from you placing the order to getting the order in your retail store or warehouse? Then you have service level. What is the probability that you will meet your customer demand without running out of stock? These three KPIs will help you understand whether your supply chain is fast, reliable, and responsive. Now let's understand a little more about these three important metric cycle time, lead time, and service level. Let's start with cycle time. In simple terms, cycle time is the time taken to complete a specific task or activity. This could be how much time does it take to unload a truck or how much time does it take to restock a fresh produce, or what's the time between picking an order and getting it to the shelf? That's your cycle time. So let's say unloading a truck, sorting it out, and shelf stocking takes around 2.5. That's the cycle time. If you bring it down to 1.5, you can save labor, speed, and availability. So how do you move your items more efficiently in a retail store? That's your cycle time. If you have a lower cycle time, you can replenish your inventory faster and you can just store them in your warehouse or in a retail store. Then you have lead time. So if you're placing a order of a product at any point in time, how much time will it take for that product to reach to your warehouse or retail store? So the total time from placing an order to receiving that good, that's your lead time. So let's say if you're ordering milk in Monday evening, it will arrive next morning. In fact, most of groceries and fresh fruit in a retail store arrives in a day every morning around six or 7:00. But on the flip side, if, let's say you are suddenly running out of stock of chocolates or spices, and if you're placing an order, they'll say, Hey, we're going to deliver in the next batch after seven days, eight days, or ten days. That's your lead time from the time you place an order to the time you receive it. It includes the supplier processing time, the transportation time, and how do you internally handle those items and put it back in the shelf? So if you have a longer lead time, you need to have a more inventory buffer, and it carries more risk. The time you have shorter lead time, it is quite responsive. So generally, that's why anytime a merchant feel that, hey, those summer season is coming, people going to demand a lot of ice cream or chocolate, so they usually store a large quantity of it. Then you have service level, which is how much demand can you meet. So what is the probability that you can fulfill a customer demand without stalking out? They try to maintain a 90, 95% service level. 98% is really difficult to match honestly. But let's say if 100 customer want bread and 98% get it, then your service level is 98%. Now for these kind of product, like perishable product, you can have a lower service level because let's say you ordered good amount of bananas and in the evening, you're running out of them. So it's fine. But usually for ng perishable item, you need to maintain a really good service level because you don't have a fear of expiry. But for bananas, fresh fruits, around evening, you might expect a low service level. So you always have a trade off. Do you maintain a high service level and end up storing a way more inventory than you actually need? Or do you maintain low service level and just be at a risk of losing sales? So a good service level is around 90 to 95%, to be honest. And a supply chain always try to find a sweet spot, high service level, low excess inventory, and accurate forecasting. So perfect. If you have learned three concepts from this video, it's speed, reliability, and cost. But when you talk about speed, reliability, and cost, what exactly does it costing us? Like, you cannot maintain all three things at once because even a fast supply chain is not sustainable and become expensive really fast. So you'll understand about the total cost of a supply chain and we'll break down each and every cost structure. 13. Total Supply Chain Cost: So now in this video, we're going to understand about all the cost breakdown in a supply chain. And let's understand this by using the same example that we were using in our last few videos. We are running a grocery store where we are selling fruits, vegetables, snacks, and few other things. And we want it to break down the usual cost structure of a supply chain. Like we are buying the product in our retail store from a supplier, then we are storing these product. And let me tell you this is the cost of just the supply chain. I mean, if you are having a retail store or a warehouse, you're paying for electricity, rent, people, machine, equipment, and a bunch of other things. So the primary purpose of this video is to help you understand the total cost of a supply chain and give you a full picture of what it takes to move product from supplier to shelf. So let's understand about the key components of total supply chain cost, and we'll just simply talk about all the major building blocks of this cost. The first cost is procurement cost. So imagine you have ten different retail store and you have centralized procurement. In that case, you might be having a team of two, three people who constantly procure from supplier at an affordable rate. So you need to pay those people, and that's your procurement cost. Then you have transportation cost. All of the supplier from which you are ordering your item, they also need to ship those item using a third party logistic network. So you have transportation cost. Then you have inventory holding cost. So if you're buying inventory of 2000 $50,000, in that case, you need to store that inventory somewhere. You need to just that money is kind of blocked. You have to sell those product. So your storage, rent, refrigeration, insurance, all of that is your inventory holding cost. Then you have operation cost. You might be paying salary to people, or you have to pay salary to people. That's your operational cost handling returns and those people packing the order, and then you have shrinkage and waste, which we already covered. You can see that this cost is adding up over time. And controlling the cost is very important because a grocery store always run on very thin margin around eight to 10%, if you do not control your cost, well, you might run out of business. So let's say if suddenly fuel prices are going up, in that case, you need to make sure that that specific cost is finally transferred to the user and you are not going below a certain level when it comes to margin. You need to reduce wastage. You need to make sure that you have a consistent footfall of customer and your supplier lead time doesn't grow. They are just delivering you the items at which they are promising and you should always carry extra safety stock and we'll understand more about safety stock where I'm going to give you some example and assignments to complete. But a good supply chain should be fast, reliable, and cost efficient. That's the primary aim of this video. How do you make smarter decisions so that your supply chain is efficient, cost effective? So to manage supply chain, you need to work on tracking the cost of each and every product that you're procuring from each supplier. You'll start talking to more supplier, and that's a job that your procurement team needs to do. So constantly, you should have at least three, four suppliers whom you can negotiate by buying in bulk. But again, you have to make sure that you're not holding too much. You can reduce the delivery time. You can have just in time delivery. You can negotiate, ask them to take the product back in case it expires or in case if it goes bad. Many suppliers do have the policy that, hey, if you're not able to sell the product, you can send it back and we can give you 80% back. Switching from daily to twice a week delivery might save you some cost because these people need much more predictable demand from you and they can ship in bulk quantity. You can do a bunch of things to save on cost, and I'm going to give you some assignment to understand this topic, much better when we move ahead in the course. 14. Types of operations systems: So great. In this video, we'll talk about the different types of operation system in a warehouse, a retail shop, a bakery. This is just for example, I wanted to make things super simple for you to understand, and that's why I'm taking examples of all the things that you see in your day to day life. Like all of us go to a retail store, we visit a bakery, a hospital, and that's why these examples are easy to understand. Some of you may not visit a warehouse, a distribution center regularly. That's why I'm not taking those complicated examples because the main purpose of this video is to make things simple for you. So let's understand about the different types of operation system. And in this video, we're going to talk about three main operation system project, batch, and continuous. But let's first understand what is an operation system anyway. So when you look at every grocery store that you see near you, it doesn't just sell things to you. It also manufactures and modify things for you. For example, so many grocery store just procure raw fruits and vegetable, and then they just simply chop it, cut it, and pack it for you so that you don't really have to do all of that yourself. And some of them also make fresh bread for you. They have some kitchen stuff which pack and cut fruits for you. They have staff that do labeling and a bunch of other processing. When you think carefully, all of this process requires system planning and organization. And when you look carefully, all of this process requires a system of planning, organization, and execution in their day to day work. An operation system is how everything works in a grocery store, and that's how they turn inputs into outputs. So now let's understand about the different types of operation system. We'll start with our first kind of operation system, project based operation. So as the name suggests, project operation is used for unique one time effort with a clear end date. So when you hear the term project, it has to be unique. It has to have a clear date and clear requirement and effort. For example, let's say a company is setting up a new retail store. In that case, there is a specific set of people who will do certain processes to set up a new retail store. Like they have to remodel the whole meat section and dairy section. They have to set up whole shelf, the walking area, where exactly the billing counter would be. They might be installing the new freezer and a bunch of other things inside a retail store and once it is up and running, then the day to day manage team will handle it. When you set up a new project with a due date with clear deadline and processes, in that case, you need to work with multiple departments. It requires a lot of planning and coordination and you have clear milestone and deadline. So this is your project based operation. Then you have batch operation. Now, as the name suggests, batch operation handles work in small or medium sized group or batches, and these could be continuous or they can happen one by one. For example, if you go to a bakery shop, they prepare a batch of every single bread or cake every single morning or maybe on an alternate day. And then they sell it in their retail shop or from the same bakery where they are producing these items. So Baker produce 100 bread of loaves in the morning. They pack 30 fruit bowl at once, and they just print all of their labels and clean all their utensil every single day. That's your batch operation. Now the system in the batch operation should be efficient and flexible and you can't really expect a really high volume in a bakery shop and they can't really adjust that demand as well because they produce the batch every morning, so if they're running out of stock, the customer has to wait for the next day. The only trade off that batch are easier to control, but you can create small wait between two different batches. You can't really have a continuous supply of these products. The third one is continuous operation where the works never stop. So as the name suggests, continuous operation are used when something runs nonstop and produce a large volume with very little variation. So if you go to a manufacturing facility that produce fruit juice or milk or eggs, in that case, these are continuous operation. So places like but there is one small concept that you need to learn. How do we measure how much work in a specific store or factory can be handled in a day? That's why in the next video, we'll talk about capacity, utilization, and throughput. These three metric will help you understand how much work or output can be produced from a single factory or a manufacturing facility. Capacity, utilization, and throughput. These three important concept we're going to learn in the next video. 15. Capacity, utilization, throughput: So perfect. In this video, I'm going to explain you these three important concept, capacity utilization, and throughput. I'll try to oversimplify these concept by giving you a super simple example of a bakery shop. Now, these three concept can also be used in the context of manufacturing so hey, everyone. Now I'm going to explain you three of the most important concept in operation planning, and that is your capacity utilization, and throput. Now for this video, I'm going to take a very simple example just to make things easier for you to understand, and we're going to discuss about a bakery shop, and we'll understand the capacity, utilization, and throuput of a bakery shop. But in real world, you can also use these three concept in a manufacturing facility as well. So let's say you have a in store bakery where you produce some item and you sell it in the same bakery. Let's say in the bakery shop, you have two ovens, three bakers, and 8 hours in a shift. I wanted to ask you one question. How many loaves of bread can you actually bake in a day in this bakery? And to answer this question, you have to understand about three main concepts capacity, utilization and Tropot. Capacity will help you understand how much can you produce in a day from this bakery. Utilization will help you answer how much of that capacity are you using right now, and Tropot is what actually comes out of the door. Let me break it down for you so that it is easier for you to understand. Let's start with capacity which simply help you answer what's your maximum potential output from this specific bakery? And that's your capacity. Now, to answer this question, let's say one oven in a day can be 100 loaves of bread. Let's say one oven can be 100 loaves per batch, and you run three batches in a day and your capacity is simply 300 loaves per day. That's the capacity maximum capacity you have in your bakery. Now if you carefully observe this data, you will realize that your capacity depends on the equipment which is oven. You just have three ovens and if an oven can produce 100 loafs, then you can produce a maximum of 300 lovs per day. It also depends on people and it depends on time and space. If you look at capacity, it has all these constraints. Equipment, people, time and space. Now, you can increase your capacity by bringing more equipment, which is on over here by bringing a few additional people and just try to run more batches. But remember, you have limited space in this bakery. So you need to find another space where you can manage the capacity in case if you see more footfall and people demanding more of the product that you're manufacturing here. Let's talk about utilization. Utilization shows the percentage of capacity that is actually being used. So in the last slide, I explain you that a maximum of 300 loves can be produced in this Bakery, but you might not be producing 300 every single day because you obviously need to sell 300 every single day. So if you are just producing 180 loves, then your utilization is just 60%. And you can calculate it by simply dividing the maximum capacity with your current volume. So 180/300 is 60% utilization. Now a low utilization can be because of multiple factor. Maybe you are overstaffed and people are just interrupting each other or maybe you do. You can't just really sell all of these loves to the customer or you have too much of equipment, or you are just being inefficient. There could be multiple reason behind under utilization of resources. 60% is in between. You should have at least 70 to 80% utilization. Going beyond that limit is pretty difficult to achieve, to be honest. In the end, you have throughput. Tropot is the actual number of finish unit that comes out at a time. To understand Tropot, you will answer questions like how many lobs were baked and ready by the end of the day and how many salad bowls were packed and labeled. Capacity shows you the maximum output that you can produce. Utilization shows you what is the actual utilization of resources that you are having in that specific facility or store location anthropod is the actual output that you are seeing every single day. Anthropot is affected by multiple factor. Let's say one of your machine breakdown in between, then your throot will reduce if your staff is not present, if you're rejected a specific quality, if you have supply shortages. So your throbot might fluctuate a bit depending on season, people, processes. So that's your throot. Now, there's one small concept that you need to understand here. And that is, how do you understand and measure what slows down the performance of throughput in a manufacturing facility or in a retail store? How do you actually identify them and solve it? To understand it, we're going to talk about process mapping and bottleneck in the next video. 16. Process mapping and bottlenecks: So in this video, we'll discuss about process mapping and bottle neck. The main purpose of this video is to help you identify what can reduce down the actual throughput that you see in your manufacturing facility, in your bakery, or just in a normal retail location. What can actually reduce down the output? And let me try to simplify this concept by taking a super simple example. If you think about anything that you do in your life, whether you work in a company or you are trying to learn a concept, have you ever felt that you're running slowly? But you cannot really figure out the reason why. To solve that problem, that's where your process mapping comes into the picture. You look at the process and try to map down each and every step and how much time is it taking to go to the next step. This is like drawing a step by step workflow to understand each part of the process. So if you're running a retail store, and let's say you have a small team that actually wash a specific fruit, they cut it, prepare it, dyes it, and they just pack it and give it to the customer. So if you have to understand what is slowing down this team, then you will look at each step of the process all the way from picking and washing the fruits to cutting them out, packing them into containers, storing labeling these and storing them in the retail store and then moving them into the shelf. So if you want to understand what is slowing your team down, then you will go through each and every step and you try to understand how much time is it taking going from this step to this step, then this step to this step, and similarly. So that is your process mapping where you map every step of your process to understand the bottleneck. And how can you spot? The main idea is that you will first spot the bottleneck and then you fix them. Bottleneck is just a delay. So bottle neck is the slowest step in your system, and it blocks the entire flow of process. Like if you see how traffic slows down at one lane bridge, that is also a kind of bottleneck. Let's understand this with the help of a small team that cuts fruit in a retail store so that they can sell it to the customer. If you look at their process, they have to wash 100 fruits per hour. They need to cut 100 fruits per hour, but if they can only pack 50 fruits per hour, well, that's your bottleneck. Because this team is slowing down the whole process. The whole process cannot be accelerated because you have one small bottle neck over here. So the process cannot go faster than the slower step. Even if you increase this to 200 fruits per hour or let's say 150 fruits per, you still have this bottleneck of packaging where you can only pack 50 fruits per hour, and that's your bottleneck. You need to either increase the capacity or the work area, or you need to put more people so that you can remove this bottleneck and the whole process will then be accelerated. So bottleneck creates cues and delays. They increase the labor cost and sometimes they also lead to lower satisfaction. So how do you actually spot bottleneck and fix them? So the good way to identify bottleneck is when you see a queue forming at a specific station or when a certain equipment is always busy while other people and other equipments are waiting for it, or when you see your staff sitting idle because they are overwhelmed with the capacity or when you see some inventory or product packing up in one specific area, let's say, in this case, there was too much veggies that was being cut and lied in a specific area because there wasn't enough people to pack those veggies. And sometimes the bottleneck are obvious. If you go to a coffee shop and you see a lot of rush early in the morning, that's obvious because everybody needs a coffee early in the morning. Or you might see a lot of rush in the checkout. These are obvious bottleneck because you have a lot of people in a specific location. So you will see bottleneck at every single space. In fact, companies do plan for this bottleneck. Like in a coffee shop, if you have so many people walking in, at least you can put more coffee machines so that you can serve coffee quickly, if not the bread or ros or any other food item. If that is taking time, it's perfectly fine, but at least you can put more coffee machine and you serve faster. So you have different things that you can do to fix the bottle neck. You can add more staff and tools at a specific step. You can rearrange the workflow so that you don't have a large queue of people. You can just redistribute people in three different queues. You can train staff members to work more efficiently at certain point in time, and you can also automate repetitive task if possible. Like, let's say, if people are disturbing the staff every single time for sugar, for spoons, you can just put them publicly at a specific location so that they can just pick it up and use it themselves. So you can do a bunch of things to reduce down the bottleneck and improve the entire system. 17. Case Study: Forecasting Supply Chain Disruptions in Manufacturing: So Hey, everyone. Now we're going to look into a case study assignment, and in this case study assignment, we're going to understand the forecasting of a supply chain disruption in manufacturing. Now the reason I'm creating videos on this case study assignment is because I want to help you understand how you can to use all the concept that you have learned in the course and solve a real world problem. So first, I'm going to explain you the problem statement, all the business challenges we have. And after that, we have a small dataset where I'm going to explain what exactly is there as a raw data and what you need to calculate. And in the end, I'm going to give you a detailed solution in case if you want to solve this assignment all by yourself. In fact, I'm going to solve half of it, and you need to solve the other half of the assignment in this case. So let's start with the problem statement and business challenges, and then we will jump into the data set and we will look into all the data we have and what we need to calculate. So, perfect. So the problem statement is that you are a product operation analyst at a Fortune 500 company, and your company specialize in the production of industrial engine and automotive machinery, and your operations span across different countries across the globe, like North America, Europe, Southeast Asia, and you are relying on just in time inventory management. This is a type of inventory management where you have a very small lead time. Over the past two quarters, you have seen recurring supply chain disruption that have led to the missed delivery SLAs, idle production lines, and also cause a customer dissatisfaction. Now, the leadership wants to have a small team that keep developing a forecasting and risk stimulation model that identifies the potential supply delays in advance and estimate their impact on the production cost and delivery timelines. You need to create this forecasting supply chain disruption in manufacturing in their supply chain resilience dashboard. Now, this is a very small assignment. Obviously, it's difficult to stimulate all the factors in the actual world, but this is going to give you a really good idea. Now, there are a couple of business challenges. The first one is the volatile delivery timelines. Your shipment schedule in many cases, are not predictable because things like port concession in China, weather events in Europe can cause a lot of delay. You have volatile delivery timelines and they're not predictable. That's the first business challenge. The second one is the lack of suppliers accountability. There are some suppliers that are consistently delaying the deliveries, but you continue to purchase and they don't really take responsibility. Third one is your demand shock. Sometime your last minute BTB orders cause the demand surge for OEMs or original equipment manufacturers for the key component, things like bearings and PCBs, because they got a sudden demand, it causes cascading delays in the whole supply chain. Fourth one is the blind spot in risk propagation. There is no real time signaling that can detect how a disruption in a specific geography or industry or a transportation mode can affect the downstream production schedule. You don't have data to back it up or update it dynamically in real time. Fifth problem is the poor visibility for disruption. Fifth one is the poor cost visibility for disruption. Anytime there is a delay in any of the manufacturing part, you exactly don't know how much that can cost the company to calculate the exact number. Now, these are some business challenges we can solve. There are many that we can't, but let's still proceed and try to solve the assignment. So this is the dataset that is given in the assignment. Let me explain this one by one, and then most probably we will solve the problem. So in column A, you have your supplier ID, which is unique. So every time you place order of a product from a supplier, this is a simple raw for that specific order. So for example, in row number five, we place order of 650 motor parts. But again, after some time, we place another order on 14th for 700 motor parts. So these are different types of product that we are placing with these different supplier, and this is our quantity. And we have almost 63 rows in the dataset, which is good enough to calculate. Then in Column B, you have your region, whether this supplier is in APAC region or in America or in Europe. Then in Column B, you have your region which simply shows whether the supplier is in APAC, America or it's in Europe. Then you have your shipment date when the shipment exactly happened, then you have your expected delivery date as promised by the supplier in the contract. Then you have your actual delivery date, which is obviously beyond the expected delivery date, maybe very rarely before the expected delivery date. Then you have your product type, order quantity, lead time predicted. That means how much of lead time your system has predicted based on the past trend. Then you have your transportation mode, whether this is coming from C, A or rod, have your global disruption flag. Anytime there is a big disruption globally, this flag will simply show you a yes and a no value. Then you have your is critical component which shows that is this component critical for manufacturing or not? Remember, whenever you see a critical component, you have to stop the manufacturing because you cannot continue a manufacturing without the critical component. While there are some component where you can continue the manufacturing and you can just put them later. So this flag shows if the component is critical for the company or not. Then you have your unit cost of that component, the production impact that can cause. So perfect. Now, you need to calculate in this assignment that every time if a delay happens, how much does that cost the company in terms of monetary value? How can you avoid it, prevent it by getting new supplier into your supply chain or by ordering from somewhere else. Now, after solving this problem, you need to use Pivet table and you need to calculate the suppliers efficiency across different regions and which supplier is causing us the most financial impact. So you can do the summarization, but let me first solve the problem. Let's start with or, in fact, I can go here and show you although technical concept that you need to understand. So perfect. At first, we need to calculate the actual lead time. So the actual lead time measures the number of days it took for a shipment to be delivered after the dispatch. And the actual shipment time kind of give you the ground of through the shipment duration because remember, this data is the expected time, and this is the actual delivery date. So what is the actual lead time? To calculate the actual lead time, you simply need to calculate the difference in number of days between the shipment and the delivery date. So if you need to calculate the actual lead time, you can simply calculate the difference between shipment date and your actual delivery date. Yeah, you can type days in the end so that it shows that you have it in number of days. That is our actual lead time, which is simply the difference between your delivery date and the actual delivery date. Then we need to calculate the lead time variance. Lead time variance is a difference between the predicted lead time and the actual lead time. And a positive lead time variant shows that the shipment is delayed. To calculate the lead time variance, you subtract your lead time predicted minus actual lead time. So here the lead time predicted was 20, the actual lead time was 25. So the lead time variance is five. That means the shipment was delayed further 55 days. In this case, it is negative one, which is a good thing, but this is your lead time variance. Then you have your is delayed flag. So is delayed simply means that if the order was actually delayed or not. So this is very simple. You have your expected delivery date and actual delivery date. If your actual delivery date is more than your expected delivery date, then the order is delayed, otherwise, it's not. That's your is delayed flag. Then you have delay in number of Ds. Now, this measures the number of Ds by which your shipment was delayed. We first need to validate if the shipment was actually delayed or not. We'll first check the flag. Now to calculate this, you first need to check if is delayed is yes, otherwise it doesn't make sense, and if so, what's the difference? And we simply calculate the difference between expected delivery date and actual delivery date. So the number of delay days were five, seven, 14, two. Perfect. Now the total order value, which is where we need to calculate the impact of the delay. And the impact of the delay can simply be calculated by simply multiplying the unit cost, which is your L two multiplied by your order quantity. So if your order quantity is 1,200, the unit cost is 45 in that case, the total order value is this much. Now the reason we calculated the total order value is because we will be comparing the total order value and the potential financial impact. Then we have suppliers reliability score, which simply help us understand if a specific supplier is reliable or not. Now, what's the way to calculate it? So supplier reliability score calculates the percentage of on time deliveries made by a specific supplier. So we will simply take a supplier ID and check the is delayed flag. So any supplier where the shipment was not delayed is a reliable supplier. So the formula is pretty simple. We will simply count all the supplier in column A, starting from A two, and we will check if there is delayed value is no. And if that's the case, we'll simply type zero and one as a binary value. Then you have your delay risk score. Which is where things get a little bit tricky. And let's understand about it. What is a delay risk score? So delay risk score is a composite metric that combines your delay severalty, global disruption context, and product criticality into a single risk framework. So this delay risk score is made up of all the variables that we have considered in this specific assignment. So first thing first, we will first check if the delay days are zero or not, because if that's the case, the delay risk score would be zero. But if you have some value in the delayed days, in that case, we need to calculate the delay risk score. The first thing that we are doing is that at first, if there is no delay, the delay risk score is going to be zero, there is no further calculation needed in this formula. Let's look into the second part, this one. We are adding 0.5 to the score if the component is critical. Remember, any critical component can hold the manufacturing. The simple rationale is that critical components, even minor delays can stall all the operation. So that's not this part. Anytime we feel that there is a critical component, we just add 0.5 to the score, otherwise, zero. Then let's understand about this part. What is this? So we are adding three to the score if the delay happens due to a global disruption. Remember, we had a flag of global disruption is critical, sorry, global disruption flag, column number G, which is your COVID or any strike or China trade war or anything. Now, these are systematic risks that requires executive level planning and supply chain strategy because every time a global disruption happens, it's going to be long lasting and it's going to be big. That's why we give it as three score, otherwise zero. In the end, this formula becomes a little bit tricky where we are converting delay number of days into smaller scale, which can be divided by two, and we are capping it at five. Let me help you understand this with the help of one example. Now the reason we are doing it is because we want to avoid extremely large score for a single long delays, and it can make the values more comparable. So for example, a 14 days delays contributes to seven if you divide it by two. But capping it at five will simply cap it at five. So for example, if your number of delays is zero, your score would be zero. But if number of delays is six, your score would be three, which is 6/2, but you're capping it at five. Then if your number of delays is ten, then 10/2, and you're capping it at five, that means your value becomes five and you're obviously adding this 0.5 year. But if your number of delays become 14, so 14/2, instead of seven, it capped to five and the score becomes much more reliable. Imagine you take your number of delays are, let's say, 50 instead of 14. In that case, your score becomes very high. Then it becomes 0.5 plus three plus 25. It becomes 2,828.5, which is too high. You will see extreme large value and that's why we are capping it at five. That's the main idea. Let's calculate this delay risk score by using the simple formula that I've shown you just now. So if your outer is equal to zero, you have zero number of delay days. In that case, you'll type zero. But if your two is yes, which is your critical component is yes, you will add 0.5. If global disruption is yes, you'll add three and you are simply calculating the score and capping it at five. That's my delay risk score. That's it. In the end, you have your potential financial impact of all the delays that are happening in your company, and the simple way to understand it is using this formula. Now, these formulas sound super difficult right now in Axl, but the reality is whenever you build a complicated software, most of these things are dynamic and your system keep fetching these values from some real time system. For example, you may not get global disruption flag values straightaway. You might be using some source to get the value or to calculate it. So to calculate your potential financial impact, which estimates the potential financial consequences if a delay happens, the first thing that you need to do is that you need to validate if there is an actual delay or not. So if P two is equal to nom, just put zero. But if there is some value in P two, which is delayed, in that case, you simply multiply the number of delay days with the production impact cost, which is this one that we calculated or sorry, this one that we already have in the row data set. If the component is critical, then you just multiply it by one, otherwise just multiply it by 0.5. That's the main idea. Delay in the critical parts result into a full cost impact, delay in non critical part, we're discounting this by 0.5. If you take this value and simply multiply this, you'll have your full potential financial impact. There you go. This is your potential financial impact. Now, you need to solve a few things by yourself and you can refer to the sold assignment as well. At first, you can see that now you have a good amount of data for how much of financial impact a supplier is causing. You can do a few things. Um, you can solve this assignment further and give me the average delay by supplier, and you can use pivot table. So just select all the dataset and create a pivot table. And after creating the pivot table, you need to give me the average delay by supplier, the reliability by the different transportation mode. Total financial impact by reason. In the end, you need to extract the shipment with higher risk score and identify the higher risk order so that you can plan your manufacturing properly. You can click on Insert and Pivot Table, and you can calculate all of these things easily. In case if you find it difficult to calculate, I have a dedicated section on pivot table where you can understand about pivot table and calculate step nine for this specific assignment all by yourself. It as a take home assignment. This is our forecasting of supply chain disruption in manufacturing and how it can have a potential financial impact in our company. 18. What is supply chain analytics: So here, everyone. Now we are starting a new section, and in this section, we will understand about supply chain analytics. So let's first understand what supply chain Analytics is. So supply chain Analytics use the data that is produced by various part of the supply chain, like procurement, manufacturing, and fulfillment. And it will improve the logistic process overall. So if you look at a normal supply chain, it starts with the manufacturing process. And once the product is ready, then you will normally ship the product to a warehouse where you will store all of your inventory, and then you will periodically schedule the delivery to a specific retail store or a supermarket. And this is the high level overview of supply chain. Now, let's understand at which part the data is being produced. Now, if you look at this supply chain at store level, you will get the sales data of a store. So let's say, if you are supplying to a retail chain or to a supermarket, then those store might be selling their product to different consumer, and they are producing a large amount of sales data. And this sales data will have details like the number of product being sold to a customer, what kind of product that they are selling and all of that. So at store level, you will have sales data. Now, whenever you are shipping the product or the inventory from a warehouse to a retail store, at this stage, you will have a delivery order. That means periodically, how much product are you shipping to a specific retail store and what's the frequency and the lot size? At this point, you will have some delivery data. So for example, how frequently a specific store is ordering some product from a specific company, and what is the lot size? Then from factory to warehouse, you will have some replenishment order data. So how quickly you need to replenish your warehouse in order to make sure that you have smooth operation in the supply chain. And at the factory level, you have production planning data. That means how much product you need to produce and what is the lead time and the cost related to the manufacturing process. So at every single stage of your supply chain, a large amount of data is being produced, and you have to use this data in order to improve your overall supply chain. So supply chain analytical tools and methods will help you understand what happened in your supply chain, why it happened, and how can you improve that process in the future. If you talk about the goal of the supply chain, the main goal of the supply chain is to make the customer happy. And the way you make customer happy is by making sure that the right kind of product is reaching the right place at the right time. So remember, the main goal of supply chain is the right item at the right place at the right time. So if we take a small example, if you look at the milk that we purchase from a retail store, it's the item, and this milk will reach to your doorstep either from a retail store or you are using some mobile app to order milk or any other product online. And then a delivery guy will ensure that the product is being on time and they are delivering the product at your doorstep. So this is the end goal of supply chain. The right kind of product is being delivered at the right place at the right time. So if I'll give you a simple example, let's start with variety. Now if you look at Coca Cola as a brand, they have different variety of product. So if you go to a normal retail shop in a village, then you can only find Coca Cola inside a plastic bottle. And you may not find Coca Cola inside a can or maybe a few other variants of Coca Cola like diet Ck. But if you go to a retail store in a city, in that case, you can find the coca cola inside a bottle, inside a can, or even the diet **** for all those people who are little health conscious. That means the variety of a product also depend on the purchasing power and the amount of data you have from that specific area. Then we have storekeeping unit, and supply chain analytics will also help you understand. Much stock you need to maintain in your retail store or supermarket of a specific product based on your previous sales data. Then we have lead time. So if you're running out of stock in your retail store, how much time will it take for that product to reach to your retail store from a warehouse? So that's your lead time management. And with the help of supply chain, you can solve the problem of variety, stock keeping unit, and lead time. So let me conclude the first video of supply chain Analytics. As a supply chain manager, you need to answer these difficult question, and data can help you simplify the answer or a logical explanation behind these difficult cis or decisions that you will take. The first one is how much inventory do I maintain in a warehouse or in a retail store. Now, these decisions can be subjective, but data can help you simplify the decision making part. When you have some sort of dilemma. Then second question would be, do I need to hire more people to handle the peak demand that I see at the time of a specific season or festival, or I can stretch the schedule of my existing employees or existing people I have. The third question could be, should I establish a new distribution center or the existing one are more than enough? Now, you can answer some of these question with the help of data, and supply chain analytics can solve the problem. But many times, if you don't have enough data and if you're not using supply chain analytics, then these can lead to more errors and losses. And that's where supply chain analytics part will come in. So in the next video, we will understand more about this topic. 19. Benefits of supply chain analytics: So, hey, everyone. In this video, we will talk about the benefits of supply chain analytics. So supply chain analytics can help organization make smarter, quicker and efficient decisions. And there are a couple of benefits that supply chain analytics can provide. The first one is the reduction in cost, and it will improve your ROI. Margins, and we will talk about it in a minute. The second benefit is that you will have a better understanding of risk and opportunities in your supply chain. The third one is the increase in accuracy in planning, and the fourth one is you can achieve a lean supply chain. Now, let's break down these four benefit one by one. Let's start with cost reduction and improvement in your margin. But before that, we have to understand about the different types of cost that you will normally see in a supply chain. The first one is the ordering cost, which is also known as the procurement cost. That means how much money you are spending when you are procuring a product from a manufacturer or wholesaler or a supplier. The second cost is the carrying cost. That means how much cost is incurred towards inventory storage and maintenance. The third one is the shortage cost. That means if you have a stockout in your retail store, what is the stockout cost and the cost of replenishment? Now let's understand the cost structure with the help of this simple diagram. And this is the income statement or the structure of income statement. So at the top, you have your net sales or the total amount of revenue you are generating as a company. Then we have COGS, also known as cost of goods SLT. And if you subtract this COGS from the net revenue, then you will have gross margin. And you can see that I've written two minus one is your gross margin. Let's say if you are selling a product at a price of 100 rupees and it will take 70 rupees to manufacture a product, then you are at a gross margin of 30%. Then obviously, if you are selling a product, you will have some operating expenses as well. So let's say you have some amount of operating expenses and maybe other costs related to maybe the employees cost and other costs like furniture and some other cost. And if you subtract your operating cost and total cost from the gross margin, you will have your net profit before text, also known as AbitA. And once you subtract your text, then you will have net profit after text. So these are all the cost that you will normally see in a business. Now, this might not be important for now, but in the finance part of this course, we had a good understanding about these different cost structure that you will see in a business. The main aim is that supply chain analytics can help you reduce your cost and improve margins. And the way you do that is by looking at all of this data. So let's say if your operating cost is increasing or if you have a higher inventory holding cost, supply chain analytics can help you reduce down that. The second advantage of using supply chain analytics is that it will give you a better understanding of risk. Now, when I talk about risk, obviously there is a root cause analysis that you can do. So let's say, if you are having a sokout of a product in your retail store, then you can simply ask three Y that why does the stokout happen? Then some of the person will say that this happened because the truck was late that was carrying the supplies. Then you will ask a further question. Okay, why the truck was late, then you will figure out that the truck was late because the driver was on leave. Then you will realize that the main problem of the stockout in my retail store was to ensure that you have enough driver to ship the product from a distribution center to a retail store. So that's the normal root cause analysis that you can do. And obviously, you can also do this type of analysis in the ERP or the supply chain Analytics, software that the company might be using. Other way by which you can understand the risk level is with the help of trend lines, and maybe you can look at the total amount of revenue forecast of your business or maybe the total expenses, and you can understand about the risk that you will have in your business, and that's your second benefit of using supply chain analytics. The third benefit of supply chain analytics is increase in accuracy in your planning. So supply chain analytics can help you answer or maybe can help you predict how much product you need to procure, how much raw material you need to procure for manufacturing, what product, and how much to produce and when to produce a specific product if you have issues related to shelf life. And we will discuss more about this in the coming few videos because we have Excel exercises where you will understand about all of these concepts by solving some complex problem. But let's talk about the fourth benefit of supply chain analytics. That is the lean supply chain. Now the first one is monitoring your warehouse, and we will understand more about it in the technology part of supply chain that how can you use RFID sensors to detect the amount of stock that is coming in and out of your warehouse and how you can automate your warehouse. But let's understand about some problems that you will face with supply chain analytics. The first problem is that not all the data is good or accessible. For example, if you are selling your product through supermarket, in that case, they might be maintaining a POS or ERP system where you can simply pull out the data or maybe you can request for data, and then you can understand. How much quantity does average person is purchasing and what all other things that the guy is purchasing apart from our product. And you can then understand a little more about the customer and the buying behavior. But if you're selling a product at a shopkeeper who is not maintaining any sort of data or system, then it is difficult to understand what kind of behavior does a customer have. The second problem with supply chain is the priortization. That means how will you prioritize between speed and cost? Which means if you use a mode of transport that is fast enough, then you might be paying a higher price. And this is a logistic problem that we will solve with the help of some example in the coming few videos. And in the end, you need some type of skill and technique to solve this problem. And obviously, over here, you need supply chain Analytics manager that will use techniques like regression or correlation to solve this problem. So this is all about the benefits and the problem with supply chain analytics. In the next few videos, we will discuss about the type of analytics and the technique that you can use in supply chain. 20. Types of Analytics in Supply Chains: So, hey, everyone. In this video, we will discuss about the type of analytics. The first one is Descriptive Analytics, and Descriptive Analytics will help you understand what has happened in the past. So if you have some past sales data or any other data, then you can understand about the things that has happened in the past. And that's why we call it as a Descriptive Analytics. Then we have Diagnostic Analytics. That will help you understand why a specific thing has happened. Then we have predictive analytics that will help you forecast what will happen on the future based on the past data. So if you have past sales data of last three to four years, then maybe you can forecast the future sales for, let's say, one, two or three months. That's your predictive analytics. And there are different techniques that we use for all of these different types of analytics. And in the end, we have prescriptive analytics. That means we will prescribe what should we do next if you are facing some problem. So we will first start with descriptive analytics to understand the past information. Then we will try to understand about the root cause of the problem that why that specific thing has happened in the past, and then we will try to predict the future possibility. And in the end, we will prescribe a specific thing that we need to do in order to avoid that problem. So these are the four types of analytics, or you can call it as a supply chain analytics. And as you know, supply chain analytics can help organization make smarter, quicker and more efficient decision. Now, let's understand about all of these four different types of analytics with the help of some example. So let's start with Descriptive Analytics. So Descriptive Analytics can help you understand what has happened in the past. And let's understand this with the help of some example. So as a company, you saw that your sales jumped to 20% and your customer complained were up by 8%. Now you have some data, but you don't exactly know what was the reason behind this. And that's where we use a Diagnostic Analytics, where we will understand that our sales was up by 20%, and our customer complaint were also up by 8%. The reason behind these two things is that there is a national holiday that have pushed our sales by 20%, and our customer complaint were also up by 8%. And the reason behind that was that three truck drivers were on vacation. And that's why Diagnostic Analytics can help you understand why does a specific thing happened. So now we know that our sales was up by 20%, and our customer complaint were also up by 8%, and these are the two reasons behind these two insight that we got from Descriptive Analytics. Now, how can we predict the possibility? That's your predictive analytics. So based on this past data, we can predict that our sales will increase by 10% next week, and because our sales will increase by 10% next week, now we need to use prescriptive analytics to ensure that we have enough truck drivers with us so that we don't get any customer complaint. So we will hire two new truck driver to ensure that we will not get more customer complaint. So we will first use Descriptive Analytics to understand what has happened in the past. Then we will use Diagnostic Analytics to understand why does that happen in the past? And then we will use predictive analytics to predict the future process. And in the end, we will use prescriptive analytics to solve that problem that has happened in the past. So these are the four different types of analytics. And if I put it in a graph, then on X axis, you have sophistication on Y axis, you have value. That means if you have a higher sophistication and higher value, then it's a better technique that can be used in supply chain. For example, if you look at Descriptive Analytics, it will give you the hindsight. That means what has happened in the past. So it's a low value event that can give you the hindsight. Similarly diagnostic will help you understand what has happened in the past, and it is also low value. But if you look at predictive and prescriptive, predictive is foresight. That means it can predict the future, and prescriptive is also foresight and it can help you solve the problem. So these are high value foresight forecasting technique, and we need to focus more on predictive and prescriptive apart from descriptive and diagnostic. 21. Introduction to Descriptive Analytics: So here, everyone. In this video, we will talk about Descriptive Analytics, and I may not be spending much time on this concept because obviously in the coming few videos, we will solve these complex problem or Excel exercises to understand about all these different types of analytics. But let's understand about Descriptive Analytics. So Descriptive Analytics will reveal what happened in the supply chain. So Descriptive Analytics is based on the past reports, and it will provide you the visibility and a single source of truth across the supply chain so that you can keep track of your day to day operation like shipment, you can detect incident, and you can also measure the performance of your operation. That's the main purpose of Descriptive Analytics. Let's look at few examples of Descriptive Analytics. First one is your Warehouse workload report. You can simply use Descriptive Analytics and you can look at the Warehouse workload report where you can see how many orders were prepared last week. So you can see that in this bar graph, you have Monday, Tuesday, Wednesday, Thursday, and how many orders were prepared last week and how many lines were there. So this is just one use case of Descriptive Analytics. Another one could be you can maintain a supply chain tower. That means you can see at which step there are shipment that are currently in transit. So obviously, you might have shipment at different stage of the process. So you can see that you have 25% shipment that are at the pickup stage. You have 24 shipment that are at the clearance stage. You have 11% shipment with order time 2% at the end clearance stage, 7% at the leaving airport stage, and you have 29% at the takeoff stage. So you can look at your supply chain tower, and this is one more use case of descriptive analytics. The third use case is the transportation route analysis. That means, what are the different routes that can be used to deliver a product to a store and how can you efficiently choose different routes in order to transport your product from one location to a different location? As you know, Descriptive Analytics provide hindsight. That means what has happened in the past. And for Descriptive Analytics, you will use some basic statistics technique like some average standard deviation. And with the help of this technique, you can simply calculate the inventory level the transportation route you need to choose, how can you allocate your resources really well? How well can you spend your money on labor? How can you manage your lead time and on time delivery report? These are all the problems that you can solve with the help of Descriptive Analytics. And for this, you just need to use some average and standard deviation and maybe variance and couple of other techniques. In the next video, let's understand about Diagnostic Analytics. 22. Understanding Diagnostic Analytics: So here, everyone. In this video, we will discuss about diagnostic analytics, and Diagnostic Analytics can help you investigate why something has happened in the supply chain. So we know that Descriptive Analytics will provide you hindsight, but diagnostic analytics will provide you insight. That means why something has happened in the past. We use Diagnostic Analytics to find the root cause of the problem. So for Diagnostic Analytics, you need to isolate a specific event within the supply chain management, and you need to ask the underlying con in order to solve that problem. Let's say if you look at the three problem that we were trying to solve in the Descriptive Analytics, we were looking at warehouse workload report, the supply chain control tower and the transportation route analysis. Let's say if you look at supply chain control tower, you have 25% shipment at the start pickpack stage, then why 25% shipment are there? What's the average time that these shipment will spend at a specific stage? You can break down a specific problem into more events that what is the average time that these material are being stuck at a specific stage, how we can reduce down their overall time and how we can optimize it further. So you can discover the anomalies in the data to better understand about a specific event. And let's understand this with the help of one example. So at the top, you have your actual time that a product will take from placing an order to making sure that the order is being delivered to the customer, and you have your maximum lead time that it will take to reach to the end user. So similarly, you can look at all these checkpoints and you can optimize your supply chain. But if something is broken, then you can quickly see that. Normally on an average, it will take two days from checkpoint three A to three B, and now it is taking three days. So the product is delayed because of the issue at three A to three B, and that is the problem. So you can find anomalies in a specific event. If something is broken, you can dig deep and understand about the actual issue, and that's your descriptive analytics. In the next video, we will understand about Diagnostic Analytics. 23. Overview of Predictive Analytics: So from the last few videos, we discussed about the Descriptive Analytics and Diagnostic Analytics. Let's understand about predictive analytics. So we know that Descriptive Analytics will provide hindsight. That means what has happened in the past. Diagnostic Analytics will provide insight. That means you can investigate why something has happened in the past. And predictive analytics will now provide you the foresight. That means you can predict what will happen in the future. And let's understand about predictive analytics. So Predictive Analytics will help a company to predict the possibility of multiple factors that can affect different business scenarios in the future. So Predictive Analytics will use Descriptive Analytics as a foundation, and you know that when you are trying to predict something for the future, then you need a large amount of past data so that you can find variation in the event. So Predictive Analytics will use Descriptive Analytics as a foundation to add more capabilities to detect and estimate a specific outcome, and it will allow industry to react according to the past data. Now to predict the retail sales, we will take multiple dataset from the past. Let's say, one of the dataset we need to take is the maximum sales quantity in the last number of days. If we have more number of data, obviously, our forecasting model will be much more accurate. So we will take the maximum sales data of the last number of days. Then we will also take the data of the stockout situation. Similarly, we will also take the data related to pricing, and then we will also take some data related to the opening and the close timing or all the days where the store was closed. And then we will try to look at the sales trend in the last number of days. And once we have all of these data, then we can build a forecasting model that can help you forecast the sales in the future. For example, you from this data, you can look at the peak sales that you had in the past. You can also correlate that peak sales with the holidays, and then you can look at at what time you had stockout and what was the pricing structure. So if you have peak sales and there is a holiday the next day, in that case, can you adjust the pricing in order to avoid the stockout? So these are all the different use cases and scenarios we can make and we can predict how a future promotion can impact store sales volume and how we can support the inventory management system to avoid the stockout situation because at the time of stockout, you're losing out on revenue and you are giving a bad customer experience. And that's why in the last video, we will understand about prescriptive analytics. 24. Introduction to Prescriptive Analytics: Here, everyone. In this video, we will understand about prescriptive analytics. So you know that in Descriptive Analytics, we will get hindsight. That means Descriptive Analytics will help us understand what has happened in the past. Diagnostic Analytics will provide insight. That means we can investigate why something has happened. Predictive Analytics provide foresight. That means we can predict what will happen in the future. And similarly, prescriptive analytics will recommend the decisions, the actions, and the strategy. Based on our forecast. Prescriptive Analytics, I think by mistake, I've written perspective, but it is prescriptive. So prescriptive analytics combines the descriptive and predictive analytics to provide a specific recommendation to reach the desired goal. So if we talk about, let's say inventory, then prescriptive analytics will recommend strategy that will aim to maximize the inventory management performance. That means you are not facing a stockout situation and you are not storing a large amount of inventory that will reduce your cash flow. And for prescriptive analytics, you will use robust software structure and you need a large amount of dataset, which the software can process and they can prescribe you or recommend you something based on the amount of inventory you have in your retail store. And prescriptive analytics will help your business, and prescriptive analytics will help your business, collaborate with logistic partners to reduce and solve problem. And you can also optimize the resource in order to reach the best efficiency. And one of the ways by which you can solve a specific problem with the help of prescriptive analytics, let's take the same problem of warehouse picking route optimization. And based on the past data or the past route you have choosed when you're transporting shipment from one location to a different location, prescriptive analytics will suggest you that what is the best picking route to minimize the walking distance. And I'm sure you have seen prescriptive analytics being used in the Google Map as well. If you go from one location to a different location, Google Map will prescribe you the best route. That means that route will take the shortest time and your ride will be much more comfortable. Also, prescriptive analytics can also be used in the warehouse value added services scheduling. That means if you have to do a series of activities in a warehouse and those activities are taking X number of time, what is the best sequence of the job execution that can minimize the cycle time. So in the end, most of the time, prescriptive analytics are linked to optimization problem where you need to maximize or minimize the objective function considering multiple constraints you may have in your business. If I summarize the video, on the right hand side, we have all the problems and on the left hand side, we have analytics technique. So Descriptive Analytics can help you understand the situation. Similarly, Diagnostic Analytics can not only help you understand the situation, but it can also help you minimize the risk because you can break down a big event into smaller chunks and you can look into the problem. Similarly, predictive analytics can provide you the visibility for planning, and prescriptive analytics can optimize the operation and it can prepare you for the future scenario. These are all the different types of analytics technique that you can use, and these are all the step by step problem that you can solve. In the next few videos, we will use some Excel data to solve some problem that a business might be facing. 25. Introduction to Demand Management: So, hey, everyone. My name is Navdeep and welcome to the section of demand management and forecasting. So what exactly is demand management? Well, demand management is nothing but a planning methodology that is used to forecast and manage the demand for a specific product or service. And a really simple example of demand management is the surge pricing that you see in any food tech platform that you use. Let's say if you're using any food tech platform like Ubereds or DoorDash, if you order any product or food items in the evening or especially at the time of dinner, you may see a surge in price or let's say, some extra charge that is applicable on that product. Well, that's because you have more demand and less supply. That means there are less people on the road who can deliver the product to your doorstep. And that is why they use this surge pricing or extra charge, especially at the time of dinner. That's a really good example of demand management from a food tech platform. But obviously in this specific course, we are talking more about how exactly demand management happens in a supply chain. So let's understand demand management from a manufacturer's perspective. Now, when we talk about a manufacturer, a manufacturer normally convert raw material into a final product. In that case, he's basically integrating the customer information into the MPC system. MPC is nothing but your manufacturing, planning and control system. A manufacturer normally get all of the raw material, then he will convert that raw material into a product based on the demand from the customer side. So if you talk about all the activities that are involved in demand management, you have converting customer order into the delivery promise and balancing your supply and demand. And when we talk about demand management, you have three important things to understand. At first, we have demand management, then we have forecasting, and then we have planning. And we have a dedicated section for forecasting. But in this video, let's talk about demand management and planning. So in short, if I summarize the video of demand management, demand management is nothing but a short term planning to keep your supplies and demand in balance. And demand planning is nothing but the analysis of consumer trend, the historical sales that we had, and the seasonality data. And we have a dedicated video in almost every single concept. And in those videos, we will understand how consumer trend can also shift your demand. How can you forecast the future sales using historical data with the help of moving average or exponential smoothing? And how can you remove seasonality from a time series data? So we will talk about all of these things in the coming videos, but for now, let's stick to the main concept. That is demand management. 26. Goals and Objectives of Demand Management: Now let's talk about the goals and the objective of demand management. I mean, as a company, why we are doing demand management at the first place. And to understand goals and objective of demand management, I have to take one example. Maybe let's take Amazon as an example. Amazon is ecommerce company, and let's understand the goals and objective in case of Amazon. The main focus of Amazon is to be a customer centric company, and that is why they have to provide the right kind of product to the right price at the right time. The number one goal of Amazon is to reduce down the fluctuation that is there from the customer side. So normally, you see that people shop a lot on holidays, especially on Christmas. And to reduce down this specific spike in the demand, Amazon normally have subscribe and save kind of option. If you normally notice yourself, you will realize that you shop a lot on weekend, and that is why they normally have this subscribe and save where they are normalizing the fluctuation from the customer side. So let's say if you subscribe to a specific product, you will get that product before time every single month. And that's one of the way they are managing their demand. Another way by which Amazon is managing demand is by using dynamic pricing. And let me give you an example to understand this. So all of us are getting our monthly salary on the first week of every single month, and that's when we have a lot more money in our bank account. And that's why we shop a lot. So Amazon always increase the pricing of their product in the first few week. And in the last few weeks, the pricing is always less than what it used to be. And that's something known as your dynamic pricing. Means your pricing will change with time, with day of the month or even within hours or second. That's the example of dynamic pricing. And dynamic pricing is there in almost every single app that you use, from a food app that you use, from Airbnb, the third demand management technique that Amazon use is by combining your prime umbrella with the best seller tech. And let me help you understand this with the help of one example. Let's say you wanted to purchase a mouse or a keyboard from Amazon. Now, if you closely observe this, you will see that there are more than 10,000 different variety of mouse or keyboard that are available on Amazon. So how exactly Amazon will make sure that you are getting the right kind of product at the right price and in the shortest duration of time? Well, one of the ways by which you can get that is by storing your top selling product in almost all the warehouses. If you go to Amazon and if you search for wireless mouse, you will see that couple of the top product will have a best seller tag. And the main idea behind that is all of their best seller product are available in all of their warehouses, and that's the idea. They wanted to ship the best quality product at the cheapest rate as fast as possible. So instead of purchasing other 99,000 mouse, you will just end up purchasing few mouse that have a best seller tag, and these product will be delivered to your doorstep within a day or two. That's your prime umbrella when combined with your best seller tech, and that's also a good example of demand management because the main focus of Amazon is to provide a good customer experience, and that is possible only when you have a good speed of delivery. So for that, you have to understand things like flywheel, prime umbrella, or best seller. Already had a video on Flywheel. So if you have more number of people buying from Amazon, you will have more seller on the platform. If you'll be having more seller on the platform in the future, then they'll be putting more and more product that will further increase the variety of the product on the platform, and Amazon will ship all those different product at the cheaper rate because now they have achieved something called economy of scale. So we already had a discussion about that specific ply wheel concept in the last few videos. Another way to reduce down the cost is economies of steel and warehousing. So Amazon will always try to store their best selling product in the warehouse so that they do not like, overburden their warehouse with inventory. Another way is the variety of the product, so they have easy on boarding, quick pickup from the reseller so that they can provide more variety of a product to the end consumer. Another technique by which Amazon do demand management is by controlling the complete product or a specific category by themselves. And that is why Amazon have recently launched a white label brand with their name called Amazon Basics. And you may have seen a couple of products like a tripod or a couple of more products with the name Amazon Basics. And the main reason is they wanted to create a category kilo, and that's why all the products that are selling large quantity on Amazon and they feel that we can provide a better product at the most affordable price, and we can also control the speed of the delivery of that specific category or product. 27. Understanding Supply and Demand: So, hey, everyone. My name is Navdeep, and in this video, we're going to talk about the supply and demand in economics. So whether you're working in supply chain or you're working in marketing or let's say in retail management, you have to understand supply and demand from the customer side. So in this video, we'll spend ten to 15 minutes understanding about the basic concept of supply and demand. Now, the supply and demand will show you the relationship between the quantity of product that you can sell at different prices and at what price consumers are willing to purchase. And let me help you understand that with the help of this simple diagram. So on XXs you have the quantity of a product that you can sell on Y axis, you have price of that specific product. So let's say at $3 of price, you can sell eight unit of a product. And let's take a product like a pizza slice. So at $3 price, you can sell eight unit of this specific product. Now, the one that you can see in the blue line is your supply curve and the one that you can see in the red line is your demand curve. Point at which both of these two intersect, this is your equilibrium point. That means at $3 of price, eight people are willing to purchase this product. That's your demand from the customer site. Now, to help you understand supply and demand, the best way to understand that is with the help of demand curve and supply curve. Let's understand demand curve. Say at $10 of price, the market demand of a product is 55 unit. But if you increase the price from $10 to $12, the market demand will reduce from 55 unit to 40 unit. On the other side, if you decrease down the price of a product from, let's say $10 to $7, in that case, the market demand will also increase from 55 unit to 75 unit. So if you decrease down the price of a product like milk, people will buy more because now they will make more dishes out of milk, and this also happens vice versa. Let's understand supply curve. Similarly, just like demand curve, let's say you are a supplier, and let's say you are supplying a product or let's say 55 units of a product at a price of $16, and then you are also making some profit. But let's say suddenly somebody will come into the market, like, let's say, government, and they will force you to reduce down the price of the product from let's say $16 to $12. In that case, it may not be profitable for you to manufacture or let's say sell that product. That's why you will also see that the supply of the product is also down. That means at $12 of price, less and less suppliers are available, and that's why the supply of this product is down. Now, whenever you increase the price of the product, now you have more supply in the market. So the supply got increased from 55 unit to 75 unit as soon as you have increased the price. Simple reason is whenever you increase the price, now the suppliers can make more money selling that product, and that is why you have more supply in the market. So as soon as you reduce the price of the milk, it will become difficult or I would say, less attractive for people to sell that product in the market. And that's why you will have less and more supply depending on the price. Now, the best way to make sure that you always have equilibrium in the market is by creating a free market economy. The time you will have intervention from the government or from someone, you will have imbalances between demand and supply. Let's say somehow you have to increase the price of this product. Now because you have more supply in the market, the supply curve will shift to the right. That means now you have more and more suppliers who are ready to sell this product in the market and because obviously the prices are high so that they can make more profit, and that is why they are increasing the supply of a product in the market. Because you have less demand in the market, it will automatically adjust in some time. And that is why as soon as there is an increase in the price, more and more sellers will be interested in selling that product, but less and less consumer will be interested in buying that product. So you will have less demand in the market. So as soon as the price will fall, the producer will also stop making that product because that may not be profitable for them, and then consumer will buy more and more of that product. So I hope you understand about the different demand and supply fluctuation that happens in the market. To understand this demand curve, I have a really interesting topic to explain. That is the elasticity of demand. Now, the elasticity of demand will help you understand how reactive your demand is as soon as you change the factors such as price or income of people. And let me help you understand that with the help of an example. So if you look at insulin as a product, so if you are a diabetic person and let's say you are purchasing insulin, that case, even if the company is increasing the price of the insulin, that doesn't going to affect how much insulin you can purchase. I mean, you still need it. It's a product which is very important for people because it's a matter of life and death. That means the price of a product like insulin will have almost zero impact in terms of demand. You will still have the same demand coming from the customer side. Now this is a good example of perfectly inelastic demand. Then you have relatively inelastic demand, and you can understand this by taking gasoline as an example. Let's say if there is a decrease in the price of gasoline from hundred dollar to $80. Even if the price decreased by 20%, you will see that there is a slight increase in the demand of this specific product. Simple reason is, people will still travel the same distance. So even if there's a decrease in the price of the product, you will see there's a slight increase in the demand, not equal increase in the demand. And that's the example of relatively inelastic demand that can be seen in case of petrol, gasoline, or any fossil fuel that is used in your transport vehicle. Then you have a elastic demand a really good example is any commodity or grocery product that we use on day to day basis. Let's say if you slightly decrease down the price of a product like bread, then you will see that there's a significant increase in the demand of that product because now people are making different dishes out of bread, and that's why even if the price is decreased down by a very small number, there's a huge increase in the demand of that product because people are now experimenting with that product. They're making different dishes. They are using this product in almost every single use case, and that's why this is a really good example of elastic demand. The end, you have your perfectly elastic demand. 28. Supply and Demand Curve Analysis: Hey, everyone. Now let's understand the different types of demand curve and how your demand curve can shift to the right and left, and then we will understand about supply curve. Normally, we think that if you increase the price of a product, the demand of that product will fall. But this may not be true in almost all the cases. So let's say if you are selling a luxury product, say if you increase the price of iPhone, the demand may not fall because people are still interested in buying that product. Or let's say, if you increase the price of a very popular or very famous product in a specific area, or let's say you're selling organic juices or anything. So at the time you increase the price of those product, the demand may not fall, and that's not always the case. And that's why we will understand how your demand curve can shift to the right and left. And let's understand that with the help of you example. So your demand curve will show you the relationship between various prices level and the maximum quantity that is purchased by the customer at those price level. And increase in the demand, that means the curve shifting to the right means the consumer is willing to purchase your product even at a higher price, and the curve will shift to the right. And this will also increase the equilibrium price. So let's say if there is an increase in the population in a specific area and you are selling, let's say, organic juices, in that case, you will always see that people are willing to purchase more as soon as you increase the price. So even after increasing the price, you will have more number of quantity that you can sell in the market. That is why your demand curve is increasing. And this is also shifting your equilibrium point. The next one is the decrease in the demand. That means the consumers are willing to purchase less of a product whenever you decrease down the price, which is totally against the rule. I mean, if you're decreasing down the price of a product, consumers should purchase more, but it also depends on the type of product that you are selling. You can see that from this specific diagram, if you decrease down the price of the product from $3 to $2, you will see that there is a less demand in the market, and demand will also reduce down from 13 units to eight units. If you closely look at these two diagrams, you will see that the supply will remain the constant, even if you are decreasing or increasing the demand curve level. And we'll understand that in a minute. But let's understand all the factors that can shift the demand curve or they can increase the demand curve. That means now you're able to sell more number of quantity, even at a higher price. And there are a few factors behind that. One of them is, let's say, if the taste of people is shifting towards a more popular product that is more expensive, in that case, again, the price of the most popular product is always higher, and that's why you have more number of people that are interested in that product. Also, let's say if there is an increase in the population in a specific area, in that case, even if you increase the price of your product, you will see that there is a more demand that is coming from the customer side. Or let's say if the income of people are increasing in a specific area or in a country or in a city, then you will also see that even at a higher price, you have more demand in the market. Let's say if the price of the complimentary product will fall, in that case, you will have more demand in the market. Or let's say if the price of a substitute product will increase, people will purchase more of your product. Let's say if the price of the coffee will increase, people will purchase more of tea. Similarly, this demand can also go down or this can also shift to the left, and the reason behind that is exactly opposite of the factors that can increase the demand curve. So now we had a discussion about all the factors that can shift a demand curve. Let's talk about all the things that can shift a supply curve. And when I'm talking about a supply curve, that means we have some supply of product in the market, and that supply might be coming directly from a manufacturer or a supplier or from a retailer. That's a supply in the market. Now, the way a manufacturer is supplying you the product is by converting the raw material into the final product. So if there is a change in the price of the input, let's say if the raw material cost will increase, obviously, the supply will decrease in the market of the finished good because now less and less manufacturer are interested in selling that product in the market because people are not willing to pay a higher price and they are not able to make profit at such a lower price. So if you have a higher cost of input or raw material, or let's say semifinished good, in that case, the supply will decrease. But on the other side, if you have a lower cost of input or raw material, in that case, the supply will increase. Then we have new technology. Let's say if a seller or a manufacturer is able to sell these products more efficiently, in that case, the supply will also increase because now they are incurring less cost because of this new technology. Now, one more factor that can shift the supply curve is change in the price of the final output. So let's say if you increase the price of cupcakes. In that case, manufacturing or producing cupcakes will become more profitable. The simple reason is now these bakeries will allot more resources in the production of cupcake, and they may decrease down the supply of donuts because they may not be able to make enough profit in donuts, and now you have higher price of cupcakes, and that's why they are producing more cupcakes. Now, one more factor that can affect the supply curve is the number of seller that are there in the market. So if you have more seller in the market, the supply will increase, and if you have less seller or fewer seller in the market, the supply will decrease. 29. Factors Affecting Demand: Now in the last video, we had a discussion about all the things that can shift your demand and supply curve. Now in this video, we will understand about all the factors that affect this demand. Now, this specific diagram will help you understand all the factors that will affect the demand or the determinants of demand. Now in the last video, we had a discussion about elasticity of demand. There are some products that are very much reactive to the pricing, while there are other products that are not reactive at all. So if I talk about products like maybe milk or bread, these are your normal grocery product. These products are very much reactive to the pricing. So if you slightly reduce down their prices, you will see that people are buying in more quantity. So the demand will increase. On the other side, if you look at some luxury product or a product like insulin, in that case, even if you increase the price of the product, there may not be a significant shift in the demand from the customer side. Now. Generally, when we look at pricing, pricing is inversely proportional to demand. That means if you increase the pricing, the demand will decrease. And when we talk about income, income is directly proportional to demand. So if you increase the income, the demand will increase. Now, let's understand all these factors of demand or determinants of demand with the help of some example. So we'll come back to the same diagram by the end of this video, but let's understand all of these individual factors. Let's start with price. So price and demand are inversely proportional, as we all know, because if you reduce down the price of a product, then demand will increase and vice versa. So if the price of the nutritional and delicious lpanso mango will fall, in that case, people will buy more quantity because now they can pour cheese the nutritional and delicious mango at a lower price. Then the second factor is income. So if people are earning more, if the income of people are high, in that case, the demand generated from them is also high. And the way people can increase their income, there are a couple of factors, by the way. So let's say if there is a more foreign direct investment in a specific country, you will see that now people are earning more because now companies are paying their employee well. Now, the stock market is booming. So if the income of people is increasing, in that case, the demand generated from that specific country will also increase. Third one is your taste, habit and preference. So if people like a product, and if the demand of that product is high, even if that brand will increase the price, you will still see a higher demand of the product. And a really good example is Apple. Apple is continuously increasing the price of the product, and people are still purchasing it. And another good example is your fashion, your beauty products. And that's why taste, habit and preference may not have a direct impact on demand when it comes to pricing. Another factors that can also affect the demand is the price of related good. And when we talk about related good, we have two different categories. We have substitute product and complimentary product. So let's say if you increase the price of tea, and the people who are drinking both tea and coffee, in that case, they will start buying more of a coffee. So increasing the price of the tea is directly proportional to the demand of coffee. The time you increase the price of the tea, the demand of coffee will increase. Complimentary product are complimentary to each other. So let's say as a government, if you impose more taxes and you increase the price of let's say, gasoline, in that case, the demand of cars will fall and vice versa, because these two are complimentary product. Then you have your consumer expectation. So let's say if the consumer feels that in the coming few months, the gold price will fall. In that case, they will postpone all of their decision of buying gold. So this is your consumer expectation. You have the number of buyers. So if you have more people in a specific area or let's say they have higher disposable income, in that case, you will see that there is a higher demand of a specific product. So number of people in an area, their disposablele income is also a factor that will decide the demand in that specific market. Then you also have these miscellaneous factors as well. I don't really want to go deep into every single one of these. The main idea is you have to understand all of these factors that affect the demand, which means you have to build a thought process like an economist, although it's not a course for understanding any economic concept as such, but you need to understand about all of these factors that affect the demand in the market and how will you make sure that all of these products are available and how will you manage the demand fluctuation and the supply fluctuation in a normal supply chain. 30. Causes of Demand Variation: So, hey, everyone. In the last video, we had a discussion about all the factors that can cause demand variation. And in that specific video, we had a discussion about things like the price of the product and income of people can really affect the demand of a product or a demand of any commodity in the market. In this video, let's understand the actual cause of demand variation. Now, one of the major cause of demand variation if you are selling a seasonal product is seasonality. So let's say if you're selling products like ice cream or umbrella in the market, in that situation, you may see seasonality because these product may cause a sudden spike or a sudden increase in the demand from the customer side, and then the supplier will order the product in large quantity from manufacturer, and then this will cause a Bullwhip effect in the supply chain. And that's one of the cause of demand variation. Another cause of demand variation is fashion. So all of these fashion product will have some trend in the market and then they will go away. And these product can also cause a demand variation, especially in the offline market. Another cause of demand variation is the change in the customer income, and we already know about this. If you have more disposable income to spend in the market, finally, you will increase the price of the product. Another cause of demand variation is the global change. So let's say if there is a trade war going on or let's say if there is a Russian Ukraine war going on, in all of these cases, your supply chain will be impacted. That's one of the reason you have demand variation. Another factor maybe if let's say company is giving more discount or if they're running some marketing campaign or let's say if they're giving promotional offers. But these are obviously the micro factors that cause demand variation. Now to understand which products are impacted due to this demand variation, we have to understand that with the help of this special two by two matrix. So broadly, we have two different type of product. We have functional product that we use in our day to day life. So all our grocery item, tissue paper, or toilet paper. These are our functional product. Then on one side, you have innovative product. So things like your computer, your mobile phone, the fashion items that you're wearing, the music that you're listening to, these are innovative product. And so here we have demand uncertainty. And here we have supply uncertainty. So functional product have low demand uncertainty because you exactly know how much food you'll be eating tomorrow and how much gas or oil you'll be using tomorrow. That's why you will see that functional product will always have somewhat stable demand. But when it comes to innovative product, you can have these spike in the demand. So let's say if there's a new trend going on on Tech Talk or let's say if there's a new computer launch, or anything that is coming in. In that case, you will have a high uncertainty in demand from the customer side. So you have your low uncertainty and high uncertainty over here. Let me take the laser pointer. So you have your low uncertainty over here and high uncertainty over here. Then you have your supply uncertainty. So you have your high supply uncertainty and low supply uncertainty. In case of functional product, you will have a low supply uncertainty because your nearby retail store, the distributor and the supplier exactly know what will be the demand for, let's say, milk tomorrow or what will be the demand for meat products tomorrow. They exactly know in case of functional product. So you will have a low supply uncertainty. But in case of products like your computer, your semiconductor, you always have high uncertainty. Now, to make sure that you are adapting really quickly with the kind of uncertainty you have, you follow these different types of supply chain. So when you have low supply uncertainty and low demand uncertainty, in that case, you follow efficient supply chain. An efficient supply chain works really well when you have economies of scale. So let's say if you wanted to produce toilet paper at the cheapest rate possible, in that case, you have to produce toilet paper in millions of quantity. And that's a really good example of efficient supply chain where you are producing product in bulk quantity so that the per unit price of the product is very less. So you are having the best capacity utilization, that means you are using all of your machine, you're producing all of these product at scale, and then you have responsive supply chain, which is there in case of innovative product. So let's say if there's a new trend that is coming on TikTok or let's say if people are wearing a new kind of t shirt or new kind of clothes, in that case, you need mass customization and you also need build to order kind of technique in your manufacturing facility. Then you have your risk hedging supply chain where the supply uncertainty is very high. And let's say if there is a sudden spike in the demand from the customer side, then you will start sharing resources of your competitor, and in that case, you are maintaining safety stock, and that's your risk hedging supply chain. Then we have agile supply chain where you're pooling resources, you're maintaining inventory. Now, in case if you do not remember the difference between functional product and innovative product, um, products like your tissue paper, toilet paper, food you eat, the gas you use, all of these are the really good example of functional product. So these products have low demand uncertainty. They can have more predictable demand. They have long product life cycle. You need to maintain less inventory of these product, and the product doesn't have profit margin. And obviously, that's one of the reasons why you're maintaining low inventory. And again, if they do not have enough profit margin, you are not going to sell different variety of these product. And even if they stock out, you're not much bothered about the customer experience in the product. In case of innovative product, these products are super expensive. They have demand uncertainty. They have higher profit margin. That's why you maintain larger variety of product. And if they stock out somehow, you may have a high stockout cost and you cannot give bad customer experience. So we already know about your functional product and innovative product. IPhone is your innovative product, and maybe toilet paper is your functional product. 31. Steps in the Demand Management Process: So now let's talk about the different demand management component and process. When we talk about demand management, you can understand demand management with the help of this diagram. In demand management, you will first collect as much data as possible, so you will collect the data of your customer, what is your sales in the last maybe one or two years and all the granular details that you need. Once you have enough data that you can feed into a data model, then you will try to forecast the future demand. So when we talk about demand management component, the first component is modeling. So you get as much data as you can from your sales team, from your social media team, from your operation team, and then you feed all of that data into a specific model that you are developing. Now nowadays, you have so many different data model like Arima, Saima, and then you will have a forecasting technique. When it comes to forecasting technique, you need three important thing. You need granularity. You need a lot more data, and then you need to fine tune that data. So in the coming section in the forecasting section, we will talk about the different forecasting technique like your moving average exponential smoothing, and that's your forecasting. And whenever we talk about forecasting, you have two important types forecasting method, you have quantitative forecasting method and qualitative forecasting method. Quantitative forecasting method is largely depend on the data, and qualitative forecasting method is largely depend on the kind of intuitions or emotions or experience you have. So in quantitative, we play with numbers, and in qualitative, we play with our intuition or let's say, our basic understanding. Now in the forecasting, the most important component that you need to understand is your time series. And we have a dedicated section for forecasting. And in that specific section, we will talk about all of these different types of forecasting technique. But let's cover the last part of our demand management component. That is your demand planning. And again, demand planning, our main aim is to strike a balance between sufficient inventory that you can maintain without maintaining a surplus. So basically, you need to strike out a balance between your demand and supply. So if you maintain more inventory in your warehouse, it will increase your inventory holding cost, or let's say it will deplete your free cash flow that you have in your company. But if you maintain less, that can provide a bad customer experience. So we maintain the optimal amount of inventory by playing around with past data, by talking to your manufacturer or let's say distributor or customer, and by using all of these different forecasting technique. It's all about all of these different components in demand management. In the next video, let's talk about what is forecasting and how can you use forecasting in manufacturing, in sales and in inventory. 32. Case Study: Price Elasticity Modeling for Airline Ticket Pricing: So hey everyone. Now we're going to talk about one small case study, and this is a small assignment in that case study. In this assignment, we're going to talk about price elasticity modeling for airline ticket pricing. Now, first, let's understand about the problem statement and business challenge. The reason I make these different videos on Kas studies and assignment is because I want you to understand how you can use the concept that you have learned in the course and solve the real world problem. So these videos, case studies, assignments, all of these things will bridge the gap between the concept that you have learned and how you can use them in solving real world problem. Let's solve a real world problem, and this is a small assignment. Before that, let's understand a problem statement, some business challenges, and what needs to be solved. After that, I'm going to give you a couple of dataset and you need to solve this problem with the help of the raw data that I'll be giving, let's do it. Let's say you're working as a pricing analyst in a regional airline company. Now the revenue team is trying to optimize the pricing strategy across different route fare classes to simply improve the profitability in the company. And they have the data like booking history, the competitor price. Now, their prices are largely static, but they want to make it dynamic and they want it to include a demand elasticity and seasonal fluctuation. Now the main goal here is that they wanted to build Price Elasticity model that simply helped them determine the optimal price for different ticket types, departure window, and they simply wanted to make more money from the airline business. And they do want it to account for competition, seasonal trend, and different price point at different days on a week. So now, the business challenges that they wanted to understand how sensitive the customer demand is to the price change. They also wanted to create a model elasticity, by different route, by different class, by different season, and they wanted to identify the price point that can maximize revenue for them. So they'll be running a couple of AB test or experiment or pricing strategy to make sure that they make more money from their business. Let's talk about all the dataset that you might find in this specific assignment, and obviously, we're going to go deeper and understand about what all we have in a row dataset and what needs to be calculated. Let's first look at the row data set that we have in this specific assignment. We have a unique flight ID for every single flight. Then you have a root, which obviously shows your origin, the destination pair. So this is the origin and this is the destination. And a route normally source shows the origin and destination peer. Then you have the date of the flight, the days to departure, how many days are remaining to the departure date. Then you have the fare class, which simply means the economy, premium, and business class seat in the flight. Then you have ticket prices, competitor prices of the same flight, the number of seat that is being sold in that flight, you have total number of seat and the season. These are the number of seats that are already sold, and then you have different type of season peak off season holiday. Now if you closely look at a very limited dataset that we have, it looks like a very simple assignment. But in reality, if you work for airline company, you might have to account hundreds of variable and the data can become super complicated and elasticity modeling is way more difficult. But this is a stimulation, and I want it to give you a super simple assignment so that you can learn the concept and also understand about it. Now you have different types of assignment that you need to solve. There are a couple of medium level problem, advanced level problem, and some real world problem. The medium level problem is that, hey, you need to calculate the price elasticity of demand for every single flight. Now, if you don't know about elasticity, I'm going to explain it in a bit, but elasticity is simply your percentage change in quantity divided by change in price, and we'll be identifying roots or fare classes with high elasticity. That means they are very price sensitive. And you need to identify the roots or fear with high elasticity or that are highly price sensitive. In advanced level analysis, we obviously optimize the revenue at different price point. We will plot a demand curve to identify the price and we will actually compare the actual and theoretical price for each segment. In the end, to solve the problem, to optimize the revenue, you need to do AV testing and price stimulation, and we're also going to solve this as well. Now before we move ahead before I start solving the assignment, let's first understand about the price elasticity of demand. This is a simple economy concept that you might have learned during your college or high school. But first let me explain it once again to you guys. Price Elasticity of demand. It measures how sensitive customer demand is to the change in price. The simple formula for this is the percentage change in quantity divided by the percentage change in price. And let's understand this with the help of a simple formula. Let's say in cell ABCD, in A two, you have your old price. In B two, you have your new price. I C two, you have your old quantity, and in D two, you have new quantity. So your Elasticity would be percentage change in quantity, which is D two minus C two divided by C two and your percentage change in price. Let's understand this with some real data. Let's say you have a flight in the next week. Their price almost three weeks back was $200. Now their new price is $250. Now when they started the flight, at that time, they sold somewhere around 1,000 ticket, and now just only 11 800 tickets are remaining. So previously they need to solve 100 ticket, but during the time duration, 200 tickets are already being sold and they now just have to fulfill, or I would say fill only 800 seats or tickets. If you calculate Price Elasticity of demand, you have to first calculate the change in the percentage change in the quantity and percentage change in the price. So percentage change in the quantity can easily be calculated, which is simply your 800 -1,000, that is 0.02 or negative 20%. Percentage change in price is your positive 25%. When you divide your negative 20% with positive 25%, your elasticity is negative 0.08. Now, what does 0.08 means here? This means that now, because the elasticity is less than one, this means that this is a inelastic demand. That means customers are not very sensitive to the price change. And increasing price a bit causes a small drop in demand, but that's fine. In fact, I can just show you the different types of elasticity of demand. Let me just Google it out. I'll not open a fancy PPT, although I've explained this topic for quite some time now, but this is your price elasticity of demand. Let me open this and let me show you this with the help of some example. We talk about price elasticity of demand. You have some of the different types of price elasticity. Let me explain this with example, but let me first show you these different types of curves that you have. Let's start with elastic demand. Every time there is a slight increase in the prices, you will see a massive drop in the demand. The price went from P two to P one, which is a slight increase, but the demand went down from D two to D one. That's your elastic demand. Because when there is a slight increase in the price, the demand goes down significantly. I'll give you the example of this as well. Then you have unitary elastic demand where your price and demand moves at a unit level. If your price increased by 10%, your demand goes down by 10%. In this case, the price increase was maybe two or 5%, but the demand goes down by 25, 30%. Then you have inelastic demand, your price increased to a large level, but your demand shifted a bit. Very small. Good example could be gasoline. No matter how much price do you increase, um, you still have to travel or you still have to drive from location A to location B, demand doesn't go down much. Then you have your perfectly inelastic demand and perfectly elastic demand, which doesn't exist in the real world. Let me take some example and understand. Let me take some example and tell you about each of these different Price Elasticity of demand. So let's start with this. At the first, you have your perfectly inelastic demand, which obviously doesn't exist and it's the vertical line, the first line that you saw. No matter how much you change the price, the quantity doesn't change. This is your perfectly inelastic demand, this one. The example is that you might have some emergency medical evacuation or last minute air travel or some family emergency. In that case, people don't really care about the price, the demand remains the same. That is the example of a perfectly inelastic demand in case of an airline. I'm explicitly taking airline as an example. Then you have your inelastic demand, which is a steep slope, this one, where a small increase small where there is a large increase in price, but there is a small reduction in demand. A good example could be all necessary or essential things like insulin, gasoline, petrol. In all of these cases, no matter how much price increase you do, you will still have some demand because people need it. Insulin is important. Gasoline is important. You still need these things in your day to day life. Obviously, it will increase the inflation, but when we talk about your price elasticity of demand, it moves them slightly. So when we talk about in elastic demand or a steep slope where your elasticity is less than one, this means that a change in price cause a smaller percentage change in the quantity demanded. So when you look at a business traveler flying between different financial hubs like New York and London, even if the price is increased from, let's say $1,000 to 1,200, the company still want to send their employees. The demand might drop a bit, but not drastically. And that's what this diagram shows. Then you have your unitary elastic demand, which is this 45 degree. If you increase the price, your demand will reduce down to the same level. In this case, the elasticity is one. That means if you look at typical budget airline like any flight from Bombay to Goa, if you increase the price by 10%, your sales drop by 10%. Then you have your elasticity greater than one, which is elastic demand or flatter slop, this one, where you increase the sorry, if you increase the price slightly, your demand reduces significantly. So if you talk about leisure travel or off season, people flying in monsoon, if prices drop $200-50, people jump and start booking the tickets quickly and the other way around as well. In the end, you have your perfectly inelastic demand, which obviously doesn't exist in the real world, which is this horizontal line. In this case, you have elasticity as infinity because at the bottom, you have zero, that means your percentage change in. Price are zero because the price is consistent. That's why your Elasticity is infinite. If you divide anything by zero, it's infinite. This is in highly competitive price sensitive routes, short haul flights, Now, these kind of demands can be seen in a highly competitive aggregator, where you keep searching for flights. So even if you find a flight which is $510 cheaper, you'll book from it. So these are your different price elasticity. So these are your different price elasticity of demand. I think I've spent too much time on explaining the concept. Let's go and solve the assignment first. Maybe, let me relate this, go to the unsold sheet, and let's understand about the data. So you have your unique flight ID. I think the data is around 200 flights is of around 200 flights. You have route where the origin is the first part and the destination is the second part. And then you have your date, days to departure, like how many days are left from this flight to departure, you have the fare class, is it premium, business economy? You have ticket prices, the competitor prices, the number of seats that are being sold and the total seat that are there in the flight and you have season and you need to first create a segment. Then you need to calculate the average ticket price, the average seats sold, the percentage change in price, percentage change in quantity, Elasticity revenue. Then you'll be running some pricing experiment like what if you run Price experiment A? What if un Price experiment B? What will be the quantity in case of A, B, then you have to calculate the revenue because obviously with every price, you'll have some quantity, you multiply price by quantity, you have your revenue A and B. Then you simply need to conclude that you run price experiment A or Price experiment B and what's the elasticity. That's the mean. So let's start solving this one by one. So let's calculate the segment first. To calculate the segment, I simply combined the root with the fair class and with the season. So I simply concatenated root, and I give a space and added E two, which is your fair class and J two, which is your season. And if you simply just press it, this is your segmentation that you want to run these experiments on. So SEADEN premium class and the season in the peak season. Let's calculate the average ticket price. Now to calculate the average ticket price for a particular segment, which obviously means that this segment is from SEA to DN, and we're talking about premium class and in the peak season, what will be the average ticket price? You simply need to calculate the average, which is your F two or the FL and you need to calculate the average ticket price for this segment. For this segment, the average ticket price is 254. Let me copy this and you'll understand there might be more than one of the segment, see, you have one, two, and probably a couple of more segment, three, four, five, six. So there are so many of the similar segment where the average ticket price is 254. In fact, it's going to be the same 254 to 54 to 54. That's my average ticket price for the specific segment. Then similarly, you can calculate the average set sold for this particular segment. There you go. I use the same formula. You have a specific segment and the number of seats sold, and you're calculating the average number of seats sold. If you just copy this, hit Control F, and just paste it, you will see that the average number of seats sold would be the same. Perfect. Let's move at the top. You have your pocentase change in price. What is the potentase change in price? Well, the formula is pretty simple. Percentage change in price is simply your ticket price, which is F two, and you need to subtract your average ticket price and divide this by average ticket price. So percentage change in price is negative 0.41. Similarly, you can calculate the percentage change in quantity. The reason we are doing that is because we simply need to calculate your Elasticity. So percentis change in price, percentis change in quantity, and the main purpose here is that you need to calculate the elasticity. To calculate the elasticity, you simply divide your 02 by two. There you go. This is your elasticity. Now, you know from this diagram, you can check the elasticity. We exactly do you lie. So in this case, it's 0.92, your elasticity, less than one is kind of inelastic demand, but less than one is very closer to unitary elastic as well. Let's go back. Let's calculate the revenue. Revenue is simply your ticket price and the number of seats sold. So ticket price is F two, which is this, and your number of seats sold is Ach two. That's your ticket price. Sorry, that's your revenue. Now, let's say you want to run different pricing experiment. Let's say you want to give a discount of 10% and charge a premium of 10%. So if you're giving a discount of 10%, you can multiply your value or your price by 0.90. If you're charging a premium, you can multiply this by 1.1. So let's say I'm giving a discount of 10% in case of pricing A. Obviously, this is dynamic in most of the cases, so I'll simply multiply my ticket price and I'll just give a 10% discount. On the flip side, you can obviously give you can obviously charge a premium of 10%, in that case, you need to multiply this by 1.1, and there you go. That's your these are your two prices. Now at two prices, you need to calculate the quantity. And the way you can calculate the quantity is by simply current number of seats sold which is at two now this is the most interesting part. How do you calculate quantity by taking the elasticity data and the price? To understand that, you need to go back to the formula or the sold part in the docs. You can actually estimate the number of seat that you will sell at a newer price just by using the elasticity data. Now, this formula explains the concept of price elasticity of demand to simply estimate how many seat you would sell if you change the ticket price, assuming that the elasticity will remain constant. So seed salt is your current or your historical quantity that is sold at a original price. Elasticity is obviously your price elasticity of demand, which is typically negative. The new price minus ticket price is the change in price. And when you divide by ticket price, it simply converts the change in price into percentage. And when you multiply your elasticity with this value, new price minus ticket price or so called change in price, it predicts the change in demand. And when you add it to one, it simply add the value to 100% or, you know, it applies the percentage change to the original quantity, and that's how you can estimate new quantity that you will sell at a newer price. I know it sounds complicated. But let me give you a very simple example to understand it. So let me help you understand this with the help of real world example. So let's say your original price was 200, your new price was 250. The number of seed that you sold at the original price was 1,000 and your elasticity is negative 0.8, which means it's inelastic. So let's plug all of these things in formula. Now, 0.25 was your percentage change in price. Your elasticity is negative 0.08 and the number of seat that you can sell at this price is 800 seat. If you raise the price to $250, you will likely to sell 80 seat, which is a 20% drop in the demand compared to the last data. And yeah, that's it. I need to first calculate the quantity A and quantity B now. Quantity A is simply the 1,000 seat that you sold, which is your I guess, H two, which is 49 in this case that you have already sold. The total seat are 180, you need to simply add your P two, which is your price elasticity and you need to simply multiply this with percentage change in price, and there you go. This is your quantity A. For quantity B, you can use the same thing. You're using Price Elasticity, a value is same, and you have the same number of seeds sold, and there you go. You have quantity B. Now you will simply calculate your revenue A and revenue B. Your quantity A and quantity B obviously depends on your new prices and they use the same elasticity and based on this, you need to calculate the revenue A and revenue B. Now you have your price A, quantity A, price B, quantity B. I think calculating revenue is the easiest thing that you can do. Perfect. Now you can see that you can make more revenue from B. In fact, I would rather create a sum than just simply coming on conclusion. The sum is 58, 58, 697. In case of revenue eight, it's 4787 021. The revenue in case of B pricing experiment is higher, obviously because we are charging premium. So yeah, that's the thing. Now, one important thing that you need to understand is the total number of seat and how many vacant seat are there. Remember, the number of seats sold are 49 here and 54 here, if you add 54 to 49, you still have a lot more number of seat that you have to sell. And that's where you need to work on the Efficiency part of it, how do you make sure that you are filling your flight or most of the seat at different price point? I think that is something that we can discuss in some other video. I'm going to create a dedicated video on efficiency. I think the best example would be airline, a hotel. I'll probably create one more assignment for that. Yeah, that's all about this video. I think you're able to understand the price elasticity and the modeling and how can you sell tickets at different price and the different quantity, and how can you calculate the revenue? But I think one major piece that we are still missing is on the number of seats that are still unsold. So we'll talk more about that. 33. Introduction to Forecasting: So, hey, everyone. My name is Navdeep, and let's start the video by understanding what is forecasting. Forecasting is a process of making prediction based on the past and the present data, and then you also use your experience, and that is why a lot of people say that forecasting is art and science because you use qualitative data and quantitative data in forecasting. And then you will compare your forecast with the actual data that you have. So let's say if you're forecasting a sales, you will first calculate your sales forecast and then you will compare that sales forecast with the actual sales data that you got after a month, and then you will see what correction are required for the next time. Now, when we talk about forecasting, people are a bit confused between the difference between prediction and forecasting. And that's why let's understand what do you mean by prediction and forecasting. Forecasting, as we all know, in forecasting, we use data of previous event and then we combine that specific data with the recent trend, and then we will come up with a future event or future outcome. That's your forecasting. A really good example of forecasting is your weather forecasting. In weather forecasting, you use your past data. Let's say the temperature of last 60 minutes or last one day and based on that specific temperature data set and looking at the trend, you will try to forecast the future outcome. In case of prediction, you will try to predict something that will happen in the future with or without prior information. If you have heard of self driving car in self driving car, you feed a specific dataset. Let's say if you wanted to predict whether the object that is coming in front of a car is a human or animal, in that case, you will train that specific dataset with thousands of images of humans and all the animals around the world. And that's how the self driving technology or the self driving car would be able to predict what kind of animal or human or what kind of animal that specific object is. I'm sure you may have one question in mind. Now, why do we need forecasting at the first place? Tell us all the possible use case of forecasting. So as we all know, forecasting is a technique of prediction of future based on the result of the previous data. And forecasting is used in almost every single industry or domain. And there are so many use cases that are there for forecasting. So a simple use case is the way Uber app calculate your fare of your RB is by using all of these forecasting technique. So Uber use a technique called as multiple linear regression where they take different dependent and independent variable, and that's how they come up with a specific fair estimate. We will talk about multiple linear regression in the last section of this specific course, but you got the point. Forecasting is there in almost every single domain that you can imagine. If you wanted to forecast the stock price of a specific stock in the stock market, I mean, for that, you can use maybe exponential moving average or weighted moving average. And we will do that in the coming videos. Other possible use case of forecasting is your recommendation engine, and the recommendation engine in Amazon is built with the help of a spatial technique called as LIFT. And if you watch my retail marketing and management course, in that specific course, I have explained how exactly the Amazon's recommendation engine work with the help of a concept called LIFT and apart from these, I think you can also use forecasting in supply chain, in logistic, as well. So you can use forecasting for better utilization of resources. So let's say if you're running a retail store or let's say if you have a warehouse and you wanted to do inventory forecasting, I mean, you can still use this forecasting technique. Then you have enhancement in the quality of management. So let's say if you wanted to hire a bunch of people or you want it to, you know, have some staff requirement for the festival season. In that case, you can do forecasting of how many people will you need based on the sales data you have from the previous period. Then you have all other possible use case, so you can just, you know, get your possible revenue or sales or all of these things with the help of forecasting. In short, there are multiple use case that are there with forecasting, and we will understand about all these different use case of forecasting in the coming videos. 34. Steps involved in forecasting: Hey, everyone, my name is Navdeep and in this video, we will understand all the steps that are required for a forecasting technique. So when we talk about forecasting, we have these five important step. In step number one, you first have to identify the problem. So you have to identify what exactly do you want to forecast from a specific dataset. So let's say if you wanted to forecast revenue of a company, then you need to find out the sales or the average unit price of the product, and that's how you can just forecast the revenue based on the sales or the per unit cost of a product. And forecasting have all these multiple use case. So you can forecast how much sales can you expect in the next month or what are the expected call volume that you can expect in a call center. So the first step is obviously, you have to define your problem first. So if you wanted to forecast something, who will need that forecast and how that forecast will be used. You have to answer these two questions. Second step is by gathering data and different companies have different data pipeline. You have to see whether your company is using whatever data pipeline you have, so whether you have all of your data in SQL database or you're using a constant time series in your cloud. So you have to build a data pipeline, and then you have to gather the data. Obviously, you can take help from your manager or your business analyst or your data analyst that is working in your company to get this specific data set. Then you have your exploratory analysis, you have to find out a specific seasonality or pattern from this specific dataset. Let's say if you're taking a time series data in that specific time series data, you have to find out whether that time series data is stationary or non Stationary, whether there is a trend component or seasonality component. We will talk about all of these things in the coming videos. But basically, you have to clean that specific dataset so that you can use your forecasting technique, which is your step number four. You have to choose a model, and there are different types of forecasting model. You have your moving average, exponential moving average, regression, Arima, SEMA, and we'll talk about all of these forecasting technique in the coming videos, and then you have to evaluate your forecast, and then you have to build a data pipeline which constantly forecast or self correct itself. And for that, you can use all of these different cloud computing platform, and they have all of their time series forecasting model. So Amazon have their own time series model, Microsoft Azure, Google Cloud, but that's more towards the data analytics and less towards the business analytics use case. Now when we talk about forecasting, there are three important type of things you have to remember. The number one is obviously your forecasting where you will predict the future outcome by looking at the past data. Then you have scenario, and in scenario, you will predict the alternative future, and scenarios are basically used by business analyst or marketer, where they try to look around different sales target or scenarios or marketing channels. And in that case, they use this scenario. Third one is your backcasting, and we will have a special video about backcasting in the next section. But backcasting is nothing but you will go 5-1 backward, and then you will try to calculate or correct your forecast going forward. And we will talk about backcasting scenarian forecasting in the coming videos. 35. Types of forecasting technique: So, hey, everyone. My name is Navdeep, and in this video, we will understand about the different types of demand forecasting. The reason we are understanding demand forecasting is because demand forecasting is super essential, especially in supply chain and operation management, because in supply chain and operation management, you need to calculate your sales projection. You need to calculate your product lead time, your product pricing, and the number of people that you require in your specific manufacturing facility. That's why to understand demand forecasting, we have two different types of demand forecast we have qualitative forecasting that use historical data to determine the future possible outcome, and then you have your qualitative forecasting. And in this, we will use survey interviews, industry benchmark and competitor analysis. So quantitative forecasting is more driven towards data, and qualitative forecasting is more driven towards your intuition or your personal experience. Can see that you have all these different types of method in qualitative forecasting and quantitative forecasting. So let's say in quantitative forecasting, you have moving average, exponential smoothing, regression, and all these different types of technique. In qualitative forecasting, you have your expert opinion, market research, focus group, historical analogy, and all of the specific data set. Reason we need both qualitative forecasting and quantitative forecasting is because you need to have both art and science. Imagine you wanted to calculate the sales forecast in a specific area or let's say in a specific country. In that situation, obviously, whatever data that you have from the last few years, that is going to be useful for you because you will use that specific data to find a specific pattern. But your forecasting or your sales forecasting also depends on the disposable income of people, the GDP of a country or the economic growth rate or all the reforms that the government is making that is why you need a combination of qualitative forecasting method and quantitative forecasting method. That is why people say that demand forecasting is both art and science. Art because you need experience in value system and science because data can help you find all of these trends or seasons or let's say, these interesting insight. Now, let's dig deep into qualitative forecasting, and then we will understand about quantitative forecasting. So qualitative forecasting. So in qualitative forecasting, you have your expert opinion, market research, forcast group, historical analogy, Delphi's method, panel consensus, and end user method. So qualitative forecasting is more based on judgment and opinion, and let's understand about one of the way or one of the method of qualitative forecasting. That is your jury of execution. So let's say a you are a business analyst in a company, and after playing around with a lot of data, you end up on a specific conclusion, but the senior executive or high level executive is not really satisfied with your forecasting. And that is why you also take their feedback on what do they think about this specific forecasting technique? So you have to mix your qualitative forecasting method and quantitative forecasting method so that you can forecast the actual data for the future. So Jury of execution is nothing but the opinion that you take from higher level executive, along with the data that you have. Then you have your salesforce composite. So all these salespeople are going in the market and they are talking to all of their customer, their retailer, or I would say, their supplier, and these people have a much better understanding of the demand in the coming month or in the coming year. That's why talking to all of your sales force or sales executive can help you understand the possible demand for the future. Then you have your consumer market survey, and this is a little expensive technique because if you wanted to do a market survey, then you have to talk to all of your customer, and then you have to take their feedback, and then you have to implement that feedback in the real world. Another qualitative method is your Delphi's method. And let me simplify this Delphi's method with the help of an example. Let's say Tesla is launching their electric vehicle into a new maybe developing country and they wanted to forecast the amount of car that they can sell in a specific country. In that case, they have to have some dataset that they can rely on, but apart from that, they also need a lot of qualitative method, they need a panel of experts that can help them understand what is the GDP growth rate, what is the average ticket size in case of car, and what is the disposable income people have? What is the per capita income? And they need to understand all of this qualitative data or all of these things like compliances, rules, regulations, texts, and all of that. And that's why they use this Delpis method to appoint a specific expert that have knowledge in a particular domain. So they might appoint expert that have a good knowledge in finance or marketing or production. And those people use their intuition or judgment or opinion along with the data. And that is the use of qualitative forecasting methods. Then we have quantitative forecasting method, and quantitative forecasting is nothing but forecasting based on the past data you have. So in quantitative forecasting, you can use techniques like moving average, exponential smoothing, regression analysis, adaptive smoothing, graphical method, econometric modeling, and life cycling modeling. In this specific course, we will understand about moving average, exponential smoothing. Inside exponential smoothing, we have three different types of exponential smoothing. We have single exponential smoothing, double exponential smoothing, and triple exponential smoothing. And then we will talk about regression, weighted moving average, and all of that. We will start by understanding your simple line that is your Y is equal to MX plus C, and then we will understand about moving average, weighted moving average, and exponential moving average. Then after that, we will understand about all of these different smoothing techniques, and then we will take all of these different independent and dependent variable, and then we will understand about simple linear regression. And to oversimplify all of these different techniques, let me take a very simple example. Let's say if your manager will just give you the sales of your company and a time period. In that case, you have to use either exponential smoothing or moving average. But if your sales manager is giving you the sales of a company and the time series, along with the marketing budget, the new product launch, and other dataset, like the number of sales representative that they are hiring. In that case, now you have more independent variable that is supporting the dependent variable that is sales and in that use case, you'll be using simple linear regression or multiple linear regression. But if I summarize the video, you can use quantitative forecasting to predict the stock price, to predict the weather condition, electricity consumption, heart rate monitoring, or total sales in a store. There are so many possible use case of quantitative forecasting. 36. Types of forecasting Model: Now let's understand how will you choose one of the forecasting technique. Now today you have so many forecasting technique based on different data model. And whenever we talk about forecasting, you first have to understand what kind of data you have. Let's say if you have just one single dimension data that is moving along with time, in that case, you can use time series forecasting. Let's say if you just have sales of a company with time or let's say you just have inventory of a product with time and you don't really have other dimension. In that case, you can use time series forecasting. On the other side, let's say if you have multiple variables, in that case, you can use Associate forecasting. Let's understand these things with the help of you example. Let's say if you have a Stationary time series data. Now, Stationary time series data doesn't have any seasonality or maybe any sort of uptrend or downtrend component. Let's say if a time series data is moving above and below its mean, that's a Stationary time series data because this time series data is not having a seasonal fluctuation, or this time series data doesn't have a uptrend or a downtrend. And in that specific Stationary time series data, you can use techniques like simple moving average, weighted moving average. And the second type of time series data we have is the trend component. Let's say if your time series data is increasing over time or it is showing you a uptrend or a down trend. In that case, you can use exponential smoothing. In exponential smoothing, you bring a smoothing cofficient or constant like Alpha that smoothen out that specific time series data. And few techniques that we will be using in this in this section will be your simple exponential smoothing and we'll be bringing a smoothing cofficient to make sure that our forecast is right. Let's say, apart from trend component, let's say if your time series also has a seasonality component. So let's say the sales of Amazon is increasing over time. But let's say, along with the uptrend, it is also having seasonal fluctuation at the time of Christmas, or let's say Black Friday. When you have uptrend and seasonality, in that case, we use a double exponential smoothing or triple exponential smoothing. And in both of these scenarios, we will bring additional smoothing constant, apart from Alpha. We will also bring a beta or maybe multiple smoothing constant. That's your time series forecasting when you just have a single dimension. What if you have multiple variables? Let's say you wanted to forecast the sales of maybe a car that you can sell in the market based on the pricing fuel prices, based on the type of disposable income people have, and you also have multiple variables. In that case, you can use regression analysis. So let's say if you're working for companies like Tesla or Ford and you wanted to calculate what will be the forecasting of my vehicle down the line in six months or a year, looking at the GDP growth rate, the disposable income people have the per capita income, maybe, let's say, the gasoline prices because that is also affecting the sales of EV manufacturer as well. 37. What is Time Series Analysis?: Hey, everyone. My name is Navdeep. And in this section, we will talk about time series forecasting. Now in this specific section, we will start with introducing you with time series data. Then we will understand about what exactly is time series and all the different components that you have in a time series data. Then we'll talk about things like your trend line, your seasonality in a time series, and then we will use these general forecasting technique before jumping into time series. We will talk about linear exponential smoothing, double and triple exponential smoothing, maybe regression analysis, Arima or Sima in the coming videos. But again, these are super complex techniques. So if you feel uncomfortable understanding any of these time series forecasting technique, you can just skip a couple of videos as well. Let's start by understanding what exactly is time series. So in our day to day life, we always try to forecast the future. So what will happen tomorrow or maybe next week or next month, or next year? So if you look at a time series data, a time series data normally have a timestem and a value. And if I give you a couple of example where you can use time series as a technique, let's say if you wanted to calculate the price of a specific stock after a month. Well, you can a normal time series forecasting. And let's say if you also wanted to calculate the revenue of your business, maybe after a year, then in that case, you can also use things like simple exponential smoothing. You can bring in a smoothing constant in your data. Now, there are countless possible use case when it comes to time series. But at the core, time series is nothing but when data is recorded. On timely basis, that data is known as your time series data, and the analysis of that specific data is nothing but your time series analysis. So let's say in column number one, you have your time, and in column number two, you have sales of your coffee. And if you plot both of these data, you will see that your time series will look something like this. That's a really good example of a time series data. Simply, let's say, if you wanted to plot maybe runs scored by a team in different years, you can also plot that in a normal time series manner. Let's say if you wanted to plot the revenue of Cisco with time, you can also do that. Well, these are the really good example of normal time series data. Let me give you a few more example. Let's say the revenue of Amazon 1995-2012 in a time series data, the airline miles data 1994-2003, that's also a time series data. So you can calculate all of these different time series data. Now let's understand the different types of time series data. At first, we have regular time series where you record all of your value in a regular interval or let's say in a uniform period of time. Then you have and a really good example is, let's say, if you're measuring temperature every single minute in a day, that's a really good example of a regular time series data. Then you have irregular time series data. And a good example is when data is collected in a non uniform fashion. Or let's say you have not defined a proper frequency of data collection. A good example of irregular time series data is when you are logging the error of your website in your database. So if you're using a software or let's say if you're using a third party tool, which logs error that is coming on your website, that can happen at any moment of time. That's a irregular time series data. If I talk about the possible use case of time series, you have anything from your capital market to science to your demand forecasting to medicine to image processing, there are so many possible use case of a time series forecasting. 38. Plotting Time Series Data: Hey, everyone. My name is Navdeep, and in this video, we will look at a normal time series data. And in this specific dataset, we have all these runs scored by a team in a specific year. Let's quickly plot this into a specific scattered chart, and then we will see how exactly a time series data will look like. But before that, we also have to compare this time series data with some average or with some mean so that you can look at the flow of a time series data and you can see that visually. And that is why I'll be calculating the average of this specific time series data. I'll simply type average, and I will press Control Shift down arrow to select all of this data in this dataset, and then I will hit Enter, and this is the mean or the average of this specific time series data. I'll simply copy this and paste it in the complete C column, and I'll paste the value. So now I have mean in the C column and a normal time series data in the column number B. Now I'll simply plot all of this data in a scatterplot. I'll go to the Insert tab and I will simply choose a scatter plot with smooth lines. If you look at this specific dataset, you can see that this is the mean that is in the orange line, and then in the blue line you have the run squad by a team. If you closely observe this scatterplot, you will see that this time series data is fluctuating above and below this mean, and this is a good example of a time series data. Just like this, with the help of Scatter Plot, you can plot any time series data that you have, whether you have a stock price of a company or let's say you have temperature on a particular day, you can plot a time series data with the help of Scatterplot. In the next video, let's understand about all the components we have in a time series data. 39. Components of time series data: Now let's talk about the different component of a time series data. But before that, let's look at a normal time series. We already know we have coffee sales with time, and this is a really good example of a normal time series data. You closely look at this time series data, you will see that this time series data have all of these three different component. You have your trend component, your seasonality component, and your noise component. Before we understand about all of these three component that will make up a time series data, let's understand the three important things that you need to understand. The first one is your granularity rule, then your frequency rule, then your horizon rule. The accuracy of your forecasting technique will depend on these three important rule. Let's start with granularity rule. So you have to forecast at a granular level. So let's say if you wanted to forecast vehicle sales. In that case, you have to be very specific with the kind of forecasting you want and what kind of data you have. Then the frequency rule, that means how frequently are you capturing your past data? So let's say if you have two different dataset in dataset number one, you have the past values that you were capturing every single R. In dataset number two, you past data where you are capturing your values every single minute. So the accuracy of data set two where you are more frequently capturing the value will be much more accurate than the data set one. Same with the horizon rule. Horizon rule will help you understand how much data you have in the past. So let's say if you are just taking the last one month data, that will be less accurate when compared with let's say if you're taking last one year data. So you have to understand about your granularity rule, frequency rule, and horizon rule, if you want it to make your forecast more accurate. Now, let's break down all of these three different components of time series data. That is your trend component, seas narraty component, and noise component. Let's start with trend component. Now, if you closely look at this time series data, this time series is increasing over time. That means this time series is showing you a uptrend. Now, trend will help you understand whether a time series is increasing over time or decreasing over time. A time series is increasing over time, that's a good example of uptrend time series. But if it is decreasing over time, that's a really good example of downtrend time series. And we will minimize this trend with the help of some smoothing constant. But that's a different topic altogether. Let's understand your seasonality. Now, seasonality, as we all know, if your time series is showing you a specific pattern over a specific period of time, that's your seasonality. If you closely look at this time series, you will see that your time series is decreasing, and then it is again increasing, then decreasing, then it is again increasing. So this time series is showing you a specific seasonal pattern that will repeat over time. And the reason is this time series have a seasonality component. In the end, you have your noise, also known as residue, and noise and residue is nothing but your random fluctuation that you see in your time series data. And then you also have your cyclicity, that doesn't repeat on a specific interval. So these are uncertain movement. These are all the three components of time series. When we talk about time series, you need to make sure that your time series is stationary and that time series doesn't have a trend or a seasonality component. If that time series have any trend or seasonality component, you always try to reduce it down or let's say, minimize it down by bringing some smoothing constant or let's say something like that, and we'll talk about that in the coming video. Let's start by taking a much more complex example. That's a simple time series that is increasing over time, and it is easily visible that this time series is having an uptrend and also some seasonality. When we talk about the time series data in case of a stock price, in that case, time series will become much more complicated. So let's understand your time series component with the help of the stock price. So let's say you have a company called Merck and Co that is listed on New York Stock Exchange, and this is the price data of this specific company. Now, when we talk about time series, you have level, that is your baseline of time series, then you have trend component, then you have seasonality, and finally, you have noise. Let's start with level. So all the red indicators that you see in this specific graph, this is your level, and your time series is going up and down above this mean or level. That's your level in a time series. Let's talk about trend component in a time series. If you closely look at this time series, you will see that you can't really see uptrend or a downtrend. And it depends on the kind of area you will zoom in. So let's say if you zoom in into this specific area, then you may see a uptrend component, but if you zoom in into this specific area, you may see a downtrend component. That means this specific time series is showing you a random trend, that can be uptrend or a downtrend, and that is why you need to understand something more in the trend component in a time series data. Time series data can have different types of trend. This can be a linear or exponential trend or a polynomial trend or a logarithmic trend, and you can see all of these things with the help of this graph. The final component you can have in your time series data is your seasonality. And as the name suggests, seasonality is nothing but the periodic up and down movement that you see in your time series data. As you all know, in a time series, if you wanted to know the accuracy of a time series technique, we go from low accuracy to a high accuracy technique. A good example of low accuracy time series where you may not have accurate value, but this will give you some more time. I mean, it is super easy to run rollover forecasting, simple average or weighted average. But these time series techniques have low accuracy, but if you take exponential smoothing, double exponential smoothing, you will have higher accuracy. So that's your accuracy scale in terms of different time series you have. 40. Components of time series data - Excel: So, hey, everyone. My name is Navdeep, and in this video, we will look into the different components you can have in a time series data. So I hope you already know about all these different components in a time series. So a time series data normally have a trend component, a seasonality component, and noise, also known as white noise. Now, trend component is when a time series is either decreasing or increasing over time, that's a trend component. Seasonality is when a time series is repeating in a specific interval. So let's say a time series is either decreasing a lot or increasing a lot in a specific time interval, that's your seasonality component. And if a time series have any small fluctuation here and there, that's your noise, also known as your white noise or residual, whatever you call it. Now, in a normal time series, it is difficult for a normal person to look at all of these three components, and that is why we will understand or I will oversimplify all of these individual component with the help of these three specific dataset that I have. So let's look at a normal time series that is stationary that doesn't have any trend or seasonality component. So I will simply select this specific dataset, and I will simply insert a scatter plot, let's say with smooth lines and marker. Now, this data shows the death of people in United States caused by flood. Now, in the first glance, if you look at this specific dataset, you might not able to observe any seasonality or trend component. And frankly, it is difficult to observe seasonality or trend component just by looking at the data visually. That is why there are a couple of analysis that we will be doing in the coming videos to find out whether a specific time series has a seasonality or trend component or not. So let's delete this. Let's plot this specific time series data set that we have and let me simply repeat the same exercise. Let me simply plot a scattered chart with Smoothine and marker. If you look at this specific scatter plot, you will see that these values are increasing over time, and this is a really good example of upward time series data. That means the sales of a company is increasing over time. So if you go 1990-1999, the sales of this specific company is increasing from $70 million to $1 billion, and that is why this is a really good example of uptrend time series data. Let's look at the third one. So if you look at this specific dataset, let me repeat the same thing again. And let me plot a simple scatter plot with markers. And if you look at this specific scatterplot, you will see that the value is decreasing over time. So this specific percentage data is decreasing from 88% to 72% when you go from 1970s to 2010. So these three different dataset shows three different parts of time series. So this is a normal stationary time series that doesn't have any trend or seasonality component. This shows you the trend component. This shows you the downtrend component. This shows you the uptrend component. This shows you the downtrend component. In the next video, let's look at a time series data that will have a combination of all these different component. 41. Plotting E-Commerce Revenue with Time Series: So, hey, everyone. My name is Navdeep, and in this video, we will look into a time series data that may have a combination of all these components. So in a time series, you have seasonal component, or maybe you have a trend component that can be a uptrend or a downtrend. So if the number is increasing over time, that's the example of uptrend. On the other side, if the number is decreasing over time, that's the example of downtrend. So let me quickly plot this specific dataset into a normal scatter plot with smooth line and marker. And let's look into this. Now, this time series data shows the revenue of Amazon 1995-2012. The column number B shows you the season. So in a year, you have four different quarter or four different season. In season number one, that is your January, February, March. Then you have your season number two, April May June, then you have season number three or quarter number three, July, August, September. And in the end, you have your quarter number four or season number four. That is your October, November and December. X and C, we have quarter count. So from one till maybe let me hit the down arrow till 68 quarter we have in this specific dataset, and then you have your revenue in column number D. So if you closely observe this specific scatter plot, you will see that there is a sudden spike in the quarter number four in every single year. This is because in quarter number four or in season number four, which is from October, November and December, you have two major festival in the United States. That is your Black Friday and Christmas. In both of these festival, people purchase a lot, and that is why you will see there is a sudden spike in the revenue of Amazon in the fourth quarter. That is your October, November and December. And that is why you will see a sudden spike in the revenue in the fourth quarter or in the fourth season, and then you will see a drop in revenue in the remaining three quarter. That is quarter number one, two, and three. Apart from this chart, you can also look at the same trend in the dataset as well. Let me mark all the four quarters that we have in our dataset. So if you closely look at the revenue or the sales of Amazon in the fourth quarter, you will see that there is a sudden spike. So the revenue is increasing to $8 million, suddenly from $4 million, in this specific quarter of 1997, the revenue got increased to $66 million from $37 million. Same, the revenue got increased to 252 from 153 and 676 from 355. And the main reason is in the fourth quarter, you have Christmas and Black Friday, and that is why you have a cyclic spike in the revenue of this specific time series data in the fourth quarter. And that is why this time series data both have a trend that is up trend and a seas narty because the value is repeating over a specific interval of time. That means this specific time series has a trend and a seas narty. Now, if you want to draw a trend line in this specific dataset, you can add a trend line. To add a trend line, you can choose moving average because this data is following a moving average, and you can just select a period four because in a year you have four quarter and then you can see that the trend line is going this way. Again, it's a trend line, and it is moving along with the trend. And in this time series data, we have a uptrend. So this trend line normally smoothen out all of this variation that you have in dataset. But this is showing a moving average trend line, but you can also create a exponential or linear depending on the dataset. But this is not a exponential growth in case of Amazon. That's why we have to choose moving average. But some companies show a exponential J curve, and that is why you can just also plot a exponential trend line as well. Now the most important thing that you need to understand in a time series is that time series should always be stationary. It shouldn't have any trend or seasonality, and that is why maybe after five or ten videos, we will understand about how can you reduce seasonality or trend component in a time series by introducing smoothing constant for trend component and for seasonality component. But before understanding those advanced concept, we first have to build a strong foundation, and that is why we are just visualizing all of these dataset so that we can understand how can you deal with time series data. Apart from that, you have one more component in time series. That's your noise, and there's nothing special about noise to plot this dataset. You can just simply go to the Insert tab and you can simply plot this specific noise and you can just click on this specific dataset and then you can just add a trend line and you can just add a linear trend line, because this value is fluctuating a lot, that is why you can see that this is nothing but noise because this is a Stationary time series, and in a Stationary time series data set, you do not have any seasonality or trend component. All you have is noise, and that's how the noise looks like. 42. Assignment: Time Series Analysis: Now this is the time you can practice whatever you have learned in the last few videos. So I'm giving you a very small assignment. You have to create a time series and moving average plot of the airline miles data that you have, and you have to discuss what is the reason behind the seasonality and the sudden decline in the airline miles data in the year 2000 or 2001. I'm also attaching the solution of this specific assignment. So in case if you wanted to know more about this, you can just download this Excel sheet and you can look at this solution as well. 43. Rollover and weighted moving average: So, hey, everyone. My name is Navdeep. And in this video, we will understand about all these basic forecasting technique that you can use in your business. So whether you wanted to forecast the sales of your business or let's say inventory in your warehouse or in your retail store, you can use this technique that is there in your screen. So in this video, we will talk about rollover, moving average, weighted moving average, and exponential smoothing. As you go from rollover to exponential smoothing, the accuracy of your forecast will increase and we'll understand that in a while. Now in the rollover forecasting technique, the forecasting for the next month is nothing but the actual sales of the previous month. And this is the easiest technique you can use for forecasting. This approach is also known as naive approach. In this approach, you just have to consider the actual sales of the previous month as the forecast for the next month. There's nothing complex in this specific technique, and you simply need to double click on it, and that's your rollover forecasting. Now, to understand how accurate a specific forecast is, you have to calculate the absolute error. So as we all know, the error is nothing but the difference between the actual sales or the actual result and your forecasting technique. In this case, it is rollover forecasting. Now, if you calculate a normal error or a normal difference, sometime you may have a positive value while other times you may have a negative value, and that is why instead of calculating a positive and negative error, we'll be calculating an absolute error. Now to calculate absolute error, I'll be using ABs in Excel and then I will hit tab to choose this specific function and I need to subtract my rollover forecast from the actual data or the actual result, and that's my absolute error. I simply need to double click on it and let me hit Control down arrow, and now I need to calculate the average of this absolute error in case of rollover forecasting. I'll type average and I will hit tab to choose the specific function, and I will hit up error key, and then I will press Control Shift up, and then you can hit Enter, that's my absolute error average in case of rollover forecasting. Another forecasting technique that you can use is your moving average. Now when we talk about moving average, you can take a three day moving average or a four day moving average, or a five day moving average or 30 day moving average, whatever you want it to take. Now, in this case, let's take maybe four day moving average. Now, if you wanted to take four day moving average, in that case, your forecast for the next month will become the average of the last four month. That means the forecast of month May will be the average of these four months. That is your January, February, March and April. So let me quickly do that. I'll simply type average I will press Tab to choose this function, and this is my four month moving average. If you wanted to calculate five month moving average or 12 month moving average, it's totally up to you. In this case, because the data is there in month, I have to take a four month moving average. But if the data is there in days or let's say in years, you can just adjust your moving average accordingly. I simply need to double click on it, and that's my four month moving average. If you hit F two, over here, then you can see the formula, and this is my four month moving average. But just like the previous technique, I also need to calculate the absolute error in case of moving average forecasting technique. So I'll be using ABS and then I'll be calculating the difference between the actual result and my moving average. And then I will double click on it. I'll go at the bottom by hitting Control down arrow key, and I will also calculate the average of absolute error in case of moving average. Cell type ABS, sorry, average and then I will just simply select all of this value. And now you can see that the absolute error in case of moving average is 1,203 and the absolute error in case of rollover forecasting is 13 35. Then there is a drop in the absolute error value. That means the moving average technique is better than the rollover forecasting technique because the absolute error in case of moving average is less than the absolute error in case of rolling or sorry, rollover forecasting. Now we have two more techniques left, weighted moving average and exponential smoothing. Let's also do that. Now, as you already know about weighted moving average, in weighted moving average, you give more weightage or more weight to the recent past data. That means if you wanted to calculate the weighted moving average for month May, in that case, you give more contribution to apparel data and less contribution to January data. So let me simply calculate this weighted moving average. I'll be using Apparel, March, February, and January, and I'm giving more contribution to the data that is there in April. So I'll be choosing APL data and let me give 40% weightage to this apparel data, let me give 30% weightage to this March data, and I will reduce it over time, obviously. I'll be giving maybe 20% to this specific data and maybe let's say 10% to January data. I'm reducing down the weightage of this specific data values again I've done a mistake. Instead of just writing 40, I have to write 0.4 because I have to multiply this. Similarly, I have to type 0.3, then 0.2. And maybe here 0.10 0.1 is nothing but 10%. Then I will hit Enter. Then I'll simply double click on it, and I'll go to the bottom and perfect. Now let me calculate the absolute error in case of weighted moving average, and the formula will remain the same. Absolute error in case of weighted moving average is nothing but the difference between your actual sales and your weighted moving average forecast. I'll double click on it. And I will hit control down Roche, and then you can also calculate the average of this. If you calculate the average of absolute error in case of weighted moving average, you simply need to select all of this data. And now you can see that the absolute error in case of weighted moving average is less than the absolute error in case of weighted moving average. That means we are improving our forecasting because the absolute error value is going down as we are going forward into all of these technique. That's all about weighted moving average, where we are giving more weight to the recent past data when compared with the later one. Now let's understand about the exponential smoothing. Now, exponential smoothing is a very complex technique, and we have to take so many assumptions over here. Now, the main idea of doing exponential smoothing is that instead of giving maybe a little less weightage to the month of January, we give maximum weightage to this specific month, and then we will exponentially decrease this specific value. One of the reason behind exponential smoothing is that your dataset may have uptrend, and that is why we are giving significant weightage to the last data, and then we are reducing this specific weightage exponentially over time. Now to use exponential smoothing, we will be using exactly the same formula that we know. So exponential smoothing or I would say the forecasting in case of exponential smoothing, that is your FT plus one is equal to your actual sales. That is your AT multiplied by your Alpha. That is your smoothing constant, plus your prior forecast. That is your FT multiplied by one minus Alpha. That is also known as your dumping factor. Let's apply all of these concepts into this specific technique. Now, the prior forecast or the Now the first value is nothing but the actual sales. This is my first value and this will act as a prior forecast in case of exponential smoothing. Now I'll be using the same formula. Exponential smoothing is nothing but your actual sales. That is this one, and this is multiplied by Alpha, that is your smoothing constant. And let me take any value of Alpha that is coming to my mind. Let's say let's take the value of Alpha as 0.4. So the value of Alpha lies 0-1. If the value of Alpha is 0.4, in that case, the value of one minus Alpha is nothing but 0.6. Let me add the prior forecast that is FT, and let me multiply this prior forecast with one minus Alpha. That is your dumping factor. And obviously, if Alpha is 0.6, one minus Alpha is nothing, uh, so if Alpha is 0.4, one minus Alpha is nothing, but 0.6. And you can simply just double click on it. Okay, I think up to perfect. Now I need to calculate the absolute error in case of exponential smoothing. The formula is quite simple, ABS and you have to find the difference between your actual sales or the actual data or the actual result and your exponential smoothing, and then you need to double click on it. And if you go down in this specific data set, you again need to calculate the average. So if you calculate the average, you will see that the value is decreasing over time. Now, if you closely observe at all the absolute error across all these different types of technique, you will see that the value is decreasing over time. So the absolute error of rollover forecasting was more and if you come from absolute error of rollover forecasting to the absolute error of exponential smoothing, you will see that huge there is a small correction, at least, if not a large correction. But the problem is 13 35 and 1103. It's still a very big difference or it's still a big absolute error in case of the actual sales and the forecasted sales. And that is why in the coming videos, we will see why we have such a large absolute error. Is there any trend in time series or is there any seasonality that is there in time series that we have to smoothen out so that we can reduce down this absolute error. And to reduce down this absolute error, we will also be using or playing around with the smoothing cofficient or smoothing constant. It's not cofficient, it's constant. We'll be working with that and let's understand about those topic in the coming videos. 44. Measuring Errors: MAD, SSE, and MAPE: So, hey, everyone. My name is Navdeep and in this video, we will understand about error and accuracy in case of time series. Now, when it comes to error, we have all these three different types of technique. We have mean absolute deviation, sum of square error, and mean absolute percentage error. Now I know these three technique may confuse you a lot, or I have seen a lot of people getting confused in all of these three techniques, and that is why I'll be taking two different dataset to help you understand that y mean absolute deviation or sum of squared error is not that reliable. And then in those days, you have to use MAPE or mean absolute percentage error. Let me simply first calculate the error or the difference between the sales or the actual sales and the forecasted sales in both of these dataset. Now, the first thing that we will do is to calculate the error or the difference between the actual sales and the forecasted sales. Let me calculate the error of both of these different dataset. To calculate the error, as we all know, error is nothing but the difference between your actual sales and your forecasted sales. I will hit Enter. I simply need to drag it forward, and that's my error. You can just copy this specific value and you can also paste the same thing over here, and it will give you exactly the same result. Now, again, error is no use for us, and that is why we have to calculate the absolute error because if you just add all of these error, you will reach to a value of zero, and that is why no matter whether the value is fluctuating above the mean or below the mean, you always try to calculate the absolute error. You make these values as positive, and that is why let's calculate the absolute error. Now, calculating absolute error is super easy. You simply have to type is equal to ABs and then you need to select this specific dataset, and that's your absolute error. You just need to drag this value forward and you need to copy this and paste it over here so that the same formula is applied in this case as well, and that's your absolute error. Now let's calculate the mean absolute deviation. That is your MAD. So let's type MAD and mean absolute deviation is nothing but the mean or the average mean is also your average, the average of this absolute error. That's your three. So mean absolute deviation is three. Similarly, in this case, mean absolute deviation is nothing but the average of absolute error. So if you select all of this dataset, you have a mean absolute deviation as ten. Now in the first glance, if you look at the mean absolute deviation in both of these cases, you will realize that the mean absolute deviation of this dataset is more. That is why the forecast of this dataset is bad and it's not that reliable because in this case, the mean absolute deviation is less, and that is why the forecast is more accurate. But remember, the base of both of these values are different. In this case, the base lies in just 20 to 50 range, and in this case, the base of the sales data lies in 200 to 500 range. That is why we cannot rely on mean absolute deviation because mean absolute deviation doesn't show you the percentage change, and that is why we have to rely on mean absolute percentage error. But before that, let's understand about sum of square error. That will give you a little better error in case you have the same base. Let's calculate the sum of squared error. To calculate the sum of squared error, you simply need to just square all of these values. So let's just simply multiply this value by itself, and then you can just drag it forward. You just need to copy the same formula over here as well, and then you need to simply calculate your SSE. That is your sum of squared error, and sum of squared error is nothing but the sum of squared error. And you simply need to just add all of these values and you simply need to just copy and paste the same stuff over here as well because we're dealing with two different dataset, and you will see that just like MAD, which is your mean absolute deviation, your sum of squared error is more in this space. But now the most interesting part that is your mean absolute percentage error because both of these two different dataset have a different base. That is why we'll be using mean absolute percentage error that will show you the change in the percentage, not in the actual value. Now to calculate percentage error, you simply need to subtract your forecast from your actual sales and then you need to divide that by your actual seals. Your actual sales is nothing but your C four minus forecast divided by actual seals, and we need to use a bracket. And that's your percentage error. Simply need to drag it forward, and you need to copy this over here as well and just drag it forward. Now, that's your E. Now to calculate MAPE, which is mean absolute percentage error. You first have to make it as an absolute percentage error. So to make this percentage error as absolute percentage error, I have to simply use the same formula. That is your absolute formula. Simply type ABS and just make these values as absolute because again, just like the difference, the value might be increasing or decreasing in terms of percentage because the time series data is fluctuating above or below the mean. I need to copy the same stuff over here as well and perfect. Now I need to calculate mean absolute percentage error. That means the average of absolute percentage error. So average of all of these values and similarly the average of all of these values. And again, the mean absolute percentage error should be in percentage. So I have to change this specific value into percentage and this specific value into percentage. And now you will see that this specific data set have more accurate forecast than this one because the mean absolute percentage error of this dataset is just 3.21. But in this case, the mean absolute percentage error is 9.63, and now you can see that the forecast for this dataset is more accurate than the forecast for this specific type of dataset. 45. Stationary vs. Non-Stationary Time Series: Now in the last few videos, we were discussing that you have to make your time series Stationary and you have to remove your seasonality and trend component from your time series. But why do we exactly need to do that? So in this video, we will understand why time series have to be Stationary. Because in the previous few discussion, we had a discussion that you have to reduce down your trend component or seasonality component from your time series. I'll give you a simple example of Stationary time series. You need to make sure that your time series stays around the mean, and it doesn't have any uptrend or downtrend component, which is practically not possible, because every time series data will either have some seasonality or maybe some uptrend or downtrend. But obviously, you need to make sure that your time series data is stationary and you need to make sure that your mean standard deviation and autocorrelation will always remain constant. And the way you do that is by removing, let's say, a trend component or a seasonality component. So this is your original time series bit trend line. Let me take the laser pointer. So this is your original time series data with a specific trend line, and you can clearly see that this is a uptrend component that this specific time series data have. And the way you make it more stationary is by removing the linear trend component that you have. This can also have exponential component as well if it is increasing at the exponential rate. So the way you can convert non Stationary time series is by removing this trend component. The simple reason is this variance that you have in your normal time series will grow over time whenever you have this trend component. So we have to de trend this time series data to fit a regression line, and then we will subtract it from the original dataset, and that's how we get a Stationary time series. Or that's how you convert a non Stationary time series into a Stationary time series. Let's talk about the different types of time series. So as we all know, you have Stationary and non Stationary time series. And let's look at mean variance and covariance in a time series data. In Stationary time series, these things doesn't change over time, so you can see that you have your mean value fixed along with the time. You variance is also fixed with time and your covariance is also fixed. In non Stationary time series, your mean will increase, your variance will increase, and your covariance will also increase. So in Stationary time series, you have a constant variance over time, and it will always return to the long run mean. While in a non Stationary time series, you may have a trend component or a seasonality component, and that's a really good diagram that can help you understand the difference between stationary and non Stationary time series. 46. Assignment: Simple Exponential Smoothing with Data Tool Pack: Now, I have a small assignment for you. You need to calculate this Simple Exponential Smoothing. And for that, you have a simple add on in Excel. You don't really have to do all of this hard work. The main purpose of making the last video was to make sure that your fundamentals or your concept are clear. You can simply either search RNs into this specific search bar in your Excel and you can just turn on this analysis toolbackO you can simply go to the file section, and then you can go to the option section, and then you can simply select ARNs, and you simply need to turn on this Analysis toolback. And then you will see this analysis tool pack below this data tap. So you simply need to open this data analysis tool pack, and you simply need to open it and you have to choose exponential smoothing. And in that case, you simply need to put the input range. That's it. And I have a label in this input range. The dumping factor is important. So instead of putting 0.2 as your smoothing constant, that is Alpha. You need to put here one minus Alpha. That is a dumping factor, and that is 0.8 or whatever value you have in mind. And let's say for output, I can choose whatever value. Let's say I want output over here, and I need a chart as well, and I need to simply hit Enter, and it will do all of this work by itself. So you don't really have to do all of that complex calculation that I told you in the last few videos. You can simply use Excel data analysis Tolbag or the exponential smoothing tool. But again, until your fundamentals are not clear, you will have a hard time understanding what's going on inside the Excel, and that was the main purpose of making the last few videos. I hope you are able to finish this assignment by yourself.