Effective data visualization for non-designers and other mere mortals | Jorge Camoes | Skillshare

Effective data visualization for non-designers and other mere mortals

Jorge Camoes, Effective data visualization

Play Speed
  • 0.5x
  • 1x (Normal)
  • 1.25x
  • 1.5x
  • 2x
13 Lessons (1h)
    • 1. Introduction

      3:58
    • 2. What is data visualization, anyway?

      5:46
    • 3. The Spanish Hostel

      6:37
    • 4. Active search for meaning

      4:16
    • 5. A Non-Linear Process

      4:56
    • 6. Chart Types

      4:37
    • 7. Sort, Rank & Proportions

      5:21
    • 8. Time Series and Relationships

      4:41
    • 9. Beyond the Single Chart

      4:41
    • 10. Design for Effectiveness

      5:42
    • 11. Functional Color

      6:32
    • 12. Resources

      1:33
    • 13. Conclusions

      2:04

About This Class

If you want to make effective and nice charts using office tools like Microsoft Excel, and have no formal graphic design skills, this introduction to data visualization is for you.

Think of it as a safety net that will prevent you from falling into the worst data visualization practices like treating a chart as a data dump, excessive makeup (3D effects, rainbow colors).

You’ll learn about data visualization tasks and questions, how to select the right chart to answer them, how to design for effective communication and how to use color.

Transcripts

1. Introduction: Hi, Thanks for your interest in this class. My name is Josh Cowboys, and I'll show you how Office uses another. Non designers can make more effective charts and communicate better with data. I live here in case you're wondering where my accent comes from. Much of what we'll discuss in this class comes from my book on data visualization titled That At Work Resistant to Like It From Time to Time. I also published eight books on my block. These are the most recent ones where I use icons to show how to make charts in excel and tableau. People can use those visual instructions, even if they don't speak English. So what is this class about? First of all, it's not about creating aesthetically pleasing visualizations. I'm sorry to say think or visualizations like this one as handcrafted, being aesthetically pleasing. These a requirement in this type of visualization because it helps attracting attention, and they are great to share on social media. But it is not the type of data visualisation we tend to expect from an office user. What we don't designers are expected to do is the boring mass production stuff. On a daily basis and using office tools like Microsoft Excel. Don't get me wrong, we do love making pretty charts. Problem is, we often don't have the right skills, tools or time, and our efforts tend to reflect that. How can we improve these and sexy mundane run off the mill visualizations, making them more effective and, if possible, not ugly? That's the question I'll answer in this class while trying to convince you that that the visualization is essentially a kitchen. Robert Not worry. I'm not telling them. You learn about the basics of the individualization for the office, including the type of questions a chart should be able to answer effectively or how to focus on a message on. While you can't ignore aesthetics, you'll see that trying too hard making pretty charts tend to have the opposite effect. I personally dislike when people tell me my charts are pretty. They sound like this because I don't know how to make a pretty chart, and neither should you. I'm serious. If the charts are pretty, that's just ah byproduct. A fringe benefit. When people don't know how data visualization works, they tend to miss the real reason why this chart is better than that chart and evaluate, um, using the categories they are familiar with, like pretty or ugly. So this is an entry level class. No prior experience with the tool is required. But working with data and making charts on a regular basis helps understanding why these concepts matter. Okay, let's talk about the glass project from my experience before, after comparison helps people become more aware of the insights to get when using each version. So we'll do that as a class project on. Hopefully, this class will help you improve the original chart, avoiding the results of over famous fresco restoration. So what am I asking you to the before going further, make a chart or select one that is representative off what you normally do. Make sure you can share it after the glass evaluated and try to redesign it based on what you've learned. Then create an image with the two versions side by side and posted comment on the changes on the result, and I promised to comment on each new entry. The class project really matters. It helps you switch to a new perspective of the individualization and helps me improve the class This is my first course, so I'm sure there is a lot of room for improvement. Your comments and suggestions are much welcome. So hope you enjoy the course on Let's Start by answering a simple question. What is the visualization anyway? 2. What is data visualization, anyway?: let's start this lesson by learning what they did. Visualization is on why we need it. It all starts to the data table that we transcribe into a visual form. We do that because the hybrid system is able to process large amounts of visual stimuli very quickly so we can take advantage of that. When using official transcription, we instantly generate relationships between data points, and that's what making sense of the data is basically all about. You'll find other definitions that includes specific goals or technology. I prefer to leave that open, and we find that a little level. Here are a few basic things that you should be careful about when visualising later. First, there must be a comparison. A single data point is bad and should be avoided. But if you don't at least at the reference line, there is no visualization. Second, there must be some potential you meaningful variation. If there is no variation or variation, is random, there's no point in visualizing it. Third, the shark mis transcribed the data faithfully. These bars are there because they look cool for some reason, but their proportions have nothing to do with the data table so it's okay to use a table to get specific data points. But if your goal is to better understand the relationships, you have to make those relationships visible. Also, since most individualization is practitioners have a romantic nature. We prefer meaningful relationships, although we do get it wrong from time to time. So forgiven, later sent, you can read the table, calculate a few metrics like sums are averages or visualize. It probably used to do all the above. But if the task requires evaluating relationships and patterns, you have to visualize them. Unless, of course, you like menial work. You're wasting your precious cognitive resources in low level tasks. To clarify many all data visualization, work includes tasks like comparing bar heights in the three D Bar chart are wondering is a value off the yellow bar in this chart is five or lower or matching color encoded categories to the legend in the line chart. We'll revisit this chart and see what we can do to improve it. So what you really should be doing is to feed your brain with stuff that pleases it or allows you to focus on high end cognitive tasks. That's why The visualization is essentially your kitchen, Robert, just like a chef that uses a kitchen Robert to prepare sources and get rid of other basic but time consuming tasks. Data visualization will pre process the data reviews later overload and for your brain for more interesting tests, like interpretation or evaluation. But you still need to know how to cook. In other words, you must have some level of the main specific expertise and again decide what is meaningful and what is irrelevant. It's important to understand the data. Visualization can speed up, but a detection, but you need context or knowledge to make sense of them or even see them. Let me give you two examples on the left. Signs off pulmonary disease are clearly visible. That's how this image is described. Anyway, I do recognize the talks, but that's not relevant. In other words, for the task at hand, what I take for signal is actually noise, and what I assume is nice is actually signal the chart on the right, on the other hand, is simpler on displace, easily recognizable presidents, and yet I have no idea what they mean. So the individualization is no magic. One that transforms data into knowledge. You can make a reasonably good chart using generic guidelines, but never underestimate subject matter Expertise Toby to it. Useful in a business are highly technical environment. The chart must perfect how the uses read and use it later when communicating with delight. With the lay audience, you have to build bridges to the existing knowledge, annotating the relevant features and explaining why they matter. A few words about the underlying data the expression that, if you're a visualization, was cleverly designed to fool you into thinking that you'll spend most of your time making cool charts. In fact, these proportions in the number of letters are about right, but they they work the other way around. You'll spend most of your time preparing and manipulating the data on only a small fraction . Actually, visualizing it in summary individualization is about finding meaningful relationships and use them to free the brain for more high end cognitive tasks. But to take advantage of these, you'll have to make sure those potential inside Sonko wasted. Whether it is it's you. Your peers are general audience. The chart must build bridges to existing knowledge 3. The Spanish Hostel: in this lesson, we'll discuss the diversity individualization landscape, A poetry book and a corporate report can share the same language but tend to be vastly different. This visualization can be a common expression to talk about a visual language, but there are countless ways to use it. You don't even have to look closely to see. The individualization is far from a margin ISS. It actually looks like a Spanish. You're still let me show you what I mean. With a bit off history during the Middle Ages, pilgrims went to Santiago de Compostela in Spain from all parts of Iberia and Europe. They had to stay somewhere. You can only imagine how erotic were those hostels along the Camino, where people from different social and geographic urgent had to share the same space, not to mention the horses and the mules. And the dogs in poetry is Aaron. French Spanish Hostel became an expression to signify metaphorical allowed gathering spaces where people share purpose but have very little else in common. What kinds off people can refining the data visualization. Also, a while ago, well, I mix, wrote a medium post. About the seven kinds are visualization people It's a light hearted text, but it's perfectly understand that data visualization is not a unified field. Subject to a simple list of rules and guidelines. It's critical to understand the data visualization. He's a visual language. Once you go beyond the equivalent of busy grammar and spell checking, there is no right or wrong way of doing it without first clarifying required skills, goals, audience profile or tools. For example, I am an Excel user that can't even draw a straight line. I often stay at this coastal, but in a dark room in the basement reserved for what Ally calls the Excel brute force is essentially, we tried to avoid examples like this and make Excel a bearable or even legitimate data visualization to going beyond the typical Excel charts is often met with some level off incl ability, and I tend to receive a lot off feedback like this. If you want to contrast the perspectives of what data visualization is all about, you must meet David McCandless on Stephan Few. David McCandless is the author. Off information is beautiful. He believes that aesthetics plays a relevant role and when communicating visually because it helps increase interest on engagement. This is one of his most conservative designs. Stephan Few prefers a more rational and effective attitude towards the individualization. This is how he believes the same that they should be visualized on day in his special way is right. A bar chart makes comparisons much simpler, although this chart types are not equivalent because they don't tell the same story. Problem is what you gain in effectiveness. You lose in interest. After all, it's just another bar chart and that is relevant in our attention economy. A piece of advice, though, If you are not a designer, it really makes more sense to have Stephan View as a reference, even when you don't agree with what he says. I made this population pyramid long ago. I like the idea, and I think it looks nice for an Excel chart. But the point is, I was trying to solve a problem. Is it possible to display a long term trend in a population pyramid? I was not trying to make a pretty chart for us. For non designers, aesthetics should be a by product of the search for an effective communication. In other words, don't try to make your shock Pretty. If you want to communicate effectively, you'll end up making elegant charts. If your chart is ugly, you're making a mistake. Elsewhere. S antibodies put it the best way. Toby, a designer when you're not a designer, is not being a designer, which means that if you make something like this, you're trying to be a designer and you are failing miserably. Although making visually creative stuff gets you more attention, that's not always the right metric. Here are a few pros and cons for off using standard visuals versus unique ones. On the plus side, stand out visuals like bar charts are very familiar. The costs of making them on depleting them is very low in The visualizations, on the other hand, may be needed for more complex data. There is obviously a new artistic dimension on personal style, requiring specific tools, techniques and skills. They are designed to attract attention under genocide. Standard visualizations risk to be overlooked because they are just one more example of a common shark type. To make them more appealing can be tempting to use can't effects like decorative three D effects. Excessive use off color. Unique visualizations require custom relearning from the audience, often without tangible benefits. The author can also be tempted to make them more appealing. The result can be aesthetically pleasing, but the message can be lost on. Obviously, the costs will be much higher. So to sum up, there is no one size fits all approach to individualization. There are useful guidelines and suggestions, but it all depends on the situation defined by the data goals, audience tools or skills s office uses. We shouldn't define as our role models in for graphics are other visualizations that I'm not aligned with our processes and skills on. We should accept the fact that if we focus on aesthetics, the results will be ugly. But if we focus on the message, aesthetics will follow. 4. Active search for meaning: It's easy to fall in love with the visuals, the creative process and the tools. But individualization is much more than these is much more than making charts. It implies a more active attitude towards the data. Let me give you a few examples. This is a chart I found on the website of the Greeks 36 office. I have bizarre, obvious. I like to use it as an example of a passive attitude towards the data because it's an absolutely useless chart. But if you use a longer time, Siri's, it gets much more interesting. Did you know that the Greeks believe or used to believe that getting married in a leap year drugs bad luck? That's crystal queer in a chart, but it's probably a religious thing. If you focus on civil marriages, the cycle becomes almost invisible on. The author also missed the entire trend. This reminded me over photo by Jill Papa, who staged it Toe illustrate a similar concept. Both the the chart on the photo sharing the reader most social as e selected pieces, the generate a version of reality. The changes dramatically as soon as we add more data. This shows that a chart is always an interpretation. Charge should be honest but is never objective in the common sense of the world. Let me give you a different example. This is the decline in infant mortality rate in Portugal, left wingers attributed to the revolution in the seventies that end it a dictatorship. While right wingers say it began long before that, this decline could be seen all over Europe. So maybe a better question would be. How does the poetry is right? Compared to the 28 countries of the European Union? As you can see, it more than doubled the future European year and rate in the sixties and remained at that level for many years. Then, around 1972 two years before the revolution, there was another care reform that impacted the infant mortality rate almost immediately. Conclusion the Revolution doesn't deserve credit for start blowing. The dean. The infant mortality rate and the dictatorship kept it too high for far too long, so no one really owns it. Liz Actor. Legitimate interpretations off off the starting data set. Obviously interpretation, and it legitimate manipulation are not the same thing on this Big was quote is not helpful . If you touch in the data long enough, it will confess, unless you are outright lying, distorting or cherry picking. You should indeed torture the data and hear what it has to say and then use your best judgment to filter the results. Jack is much more to the point. A graphic is no longer drawn once and for all. It is constructed on reconstructed until all the relationships which lie within it have been perceived. This is how we should approach the data just because there is an obvious pattern. It doesn't mean that there is nothing more to learn from the data, and often you obvious pattern can be misleading. So the major tech ways here is that data visualization is always an an interplay between those terms, the data on individuals. That means also seeing the later from multiple multiple perspectives on Don't worry, there is nothing wrong in touching the data. Just remember that, like any other form of communication, a chart is an interpretation 5. A Non-Linear Process: individualization. You don't make charts, you seek answers. Visualisation can display meaningless but aesthetically pleasing patterns. I believe that still counts as the individualization, or we can apply a conditional formatting to the whole table and see what happens as I read here. But in most cases, we do want to make sense of the data, and to that, we ask questions on the charts to be able to answer them. Finding those answers is a process. Here is how it happens. The arrows mean that the process is not linear. It all starts with a question on what is interesting about questions is that to some extent they reveal what you know better, and more precise questions tend to be a sign over better understanding for each question you. Soon there is a data source that you can use to get the answers so you get the data. As they say, no battle plan survives contact with the enemy, and that much can be said about our questions. New data will probably trigger new questions on data availability may force you to change your overall goals. Take a look at this chart. It displays data availability by country and here. So it's a bit matter because it's data about data. As you can see, there are many missing points on the left corresponding toe. The West countries on the right. It's mostly African countries and Pacific islands. I knew there would be some gaps, but I was overly optimistic about the data. So I needed to change my goals from a global analysis toe. A more modest one focused on European countries only. Andi, please note that I'm talking about data availability, not quality. That's a whole new discussion. Okay, Now that you have the right data, you should do your best toe. Make sure Discovery is actually discovery on not only confirmation. Try to analyze the data from other points of view, not old patterns. Not all insights are relevant, so they must be evaluated on prioritized. And then those insights must be communicated in a message crafted considering the audience . So how do you craft that message? And specifically, how do you choose the right chart? Glad you ask. Let me give you an example. Here are monthly food prices, pricing access in Hungary Since 2000 and five, this is an example of what we call a spark it'd chart. It is so confusing that young Lee insights we can get from it a generic and vague. I take too much time to be read, so how can we improve it? As a general guideline, we could say that the best chart is the one that answers a precise question, minimizing the brain's cognitive load in a specific situation. Yeah, I know this sounds like a glorified under confusing where you're saying it depends, But let's focus on the first section answering a precise question. For example, this chart could answer a question like Do foot prices and in angry display, some sort off seasonality are We could add context toe evaluate, variation. As you can see now, the data is exactly the same, and it's still a line chart. But now we have a clear question. In other words, if the question is clear in your mind, that's the first step toe. Answer it. Asking good questions is not as simple as it sounds. If you have friends in market research, ask them about questions suggested by their clients. I'm sure they will. There will be happy to share. We deal many cringeworthy examples anyway I suggest you take a few minutes reading about question design in a survey and book. Not every advice applies to our data visualization process, but this will help you become more aware off the importance off clarity and avoid ambiguity when questioning the data. Good questions to start with our about total volume. Major players, new players, how things change in space are over time. Then try answers like what is normal and what is not on how this problem relates to that problem. So that equates for this lesson are the visualising data is a non linear process, both for discovery and for communication. If the debt is new to you, started simple engineering questions on test variations and detail. Not forget that when you have were questions, selecting a chart type to answer them become much easier. 6. Chart Types: Now that you have a clear question, it's easy to find the right chart to answer it. You can choose a line chart, a pie chart, a bar chart, right? No, not really. Defining a chart by geometry like X Elders is the wrong approach. Here is a great reference to hundreds off chart types, and you'll see people suggesting new types almost on a daily basis. How can you possibly know all of those chart types You can't. What I mean is that the traditional classifications off chart types based on their geometry is no longer valid. The Locusts articulator a tool from Microsoft Research. You can use any shape and go dating to some off their properties. If Allah a new chart type, you can call it a chart, but it doesn't make sense to call it the chart type. Traditionally, we have a handful of chart types associate ID with tasks like using lines for time, Siri's or bar charts to compare data points. But what is snail chart or lollipop chart? It means very little. It's easy to manage this if you recognize that all charts are scatter plots at heart. Let me show you when you insert a scatter plot in PowerPoint, you get a template like this. It can easily be turned into a bar chart. Oh, no line chart. The original data points are still there, but we chose different shapes because some basic shapes arbiter lined with some tasks. Let me show you a few more complex examples to reinforce this idea that all charts are scattered parts of heart and that has practical consequences. These examples were all made with a scannable, charting excel. Keep that in mind. The bar chart is the natural candidate to test inside year. The Step chart uses color twinkle change blue for increase in red for the crease. The square four full plot is a visual matrix of two rows and two columns. In this example, we think data exercise and diet A would be the right combination for weight loss. This time cycle displays about for each seat. It's not just the texture for each slice. This is the bullet chart on alternative so gorgeous proposed by Stephan Few. There are several ways of making it in excel, but this one is more flexible, and it's easier tohave intermediate data points and alert's this is a simple pictograms. Chart on this is the one I prefer the most. The drunken goes. It encodes the time series into a pointer so that you don't waste precious real estate with a single data point, and you can use more than one point er, although I advise you not to use more than three. I admit I was lazy and used the donut chart for the scale, but it could be done adding a new series to the scatter plot. So my point here is that you can encode data into the visual properties off many objects. The traditional mastication off chart times is no longer helpful and can actually arm your analysis. If the analysis requires a more complex visual approach, you'll have to find the right balance between complexity and familiarity. Some tax enemies recognized that the task should be the first variable to classify charts. This example from the Autograph Gallery shows precisely that, so we should stop defining chart types based on their geometry and focus on the type of tasks do you want them to perform. For example, a regular line chart will perform poorly when displaying data with strong seasonality. Also no, no 1 to 1 relationships chart type can be used for different tasks and when tasks can be performed using multiple alternative tasks. So the major takeaways here is that when you have were questions, finding charts that answer them become easier. If you don't have a clear question, you'll be tempted to turn the chart into a data them where the audience needs to fish for what it is looking for. So it's the question that drives chart selection. That selection is not based on geometry, but on specification off tasks. Geometry can be used to identify a chart, but not as a primary classifications key. 7. Sort, Rank & Proportions: Okay, Now we know that classifying charts by the German trees of El Idea and that the type of questions is a much more interesting perspective. Let's start by questions related to sorting, ranking on distribution. The focus for these questions is mainly the individual data point. First, you want to perform simple comparisons like which are the most populous countries? How do they rank? What's the distance between the second and the third? How does my country compare? So ranking and sorting on direct pair wise comparisons are the most basic tasks. The bar chart is the traditional chart for such tasks, but there are other that you can use, like the plots are slow. Plus, let me exemplify with some historical examples. This is the first bar chart made by William Play, for I have to say it's better than many bar charts I see today, for example, many current charts display the bars sorted alphabetically, a silly mistake. The Playfair avoided comparisons are very precise when using bar charts, but there is a strong possibility that your audience finds them boring. There are a few things that you can do to minimize that try to play with alternative designs without losing its major advantage. Also, bar charts are often very inefficient when it comes to screen real state management. As you can see in these examples, much off the plate area is empty. You can use it for the title comments or even a small chart. Otherwise, you could consider other chart types. The part of a whole announces its simple on off, useful to use as a reference point before diving into more complex details. Please note that the part of a whole analysis should be exactly that. Comparing a slice to the whole, it's easy to start comparing categories without even realizing it. Try to make sure you need this type of analysis on not sorting on drinking well in play Fair is also credited for creating the pie chart, and now I am supposed to take a moment for the obligatory pie chart bashing. This is a copy off a famously bad pie chart found on the Wikipedia. Most people will tell you there are way too many slices in this pie chart. I couldn't disagree more, since this is a question that state level in the US, it's okay toe have as many slices as states. Yes, this is a terrible chart, but not because it has too many slices. Actually, too many data points is rarely the true reason why visualization fails. This chart fails because the author committed one off the worst mistakes in data visualization. Treated the chart as a data dump. Here is a better alternative. Display the four regions and add another ring with the details at state level. Very good. Label a few relevant slices and rotate a chart to somewhat refer to geography. There is no shortage off interesting options if you know what you want to say. Displaying the distribution off data points along an axis helps you to locate the general area where most data points will fall. How compact or white that area is, or if there are data points that depart from the central areas significantly. Here is the population density by state again. As you can see, most states have a population density clearly below 50 inhabitants by square kilometre. Then you have a second group of states between 50 and 100. The second chart shows that the denser ones are not the more populous, not that often. We don't have to identify all the points, and in those cases, a street plot is a good alternative to a bar chart. The problem with displaying the distribution of data points along an axis is that it generates lots of nice sensitizing distribution without losing key features is always challenging. It usually involves segmenting the axis and count the number of data. Points in each segment are being, or you can read it the other way around. Define how many later points you weren't in each beam and see where the cutting points are . That's how the box part is calculated to beans with equal number of data points for beans with equal number off data points. Then you have to define the place beyond which data points are considered out. Liars. I don't know about you, but I remember having a hard time understanding the concept off a box plot before realizing it's just getting points along the axis just like these two gram. The difference is that it is a gram. You pretty fine getting points to get frequencies. Andi, the box pot. You pre defined frequencies to get the petting points. Now you can remove the data points from the center of the distribution and leave the outliers. Suddenly, this room is probably more intuitive to read, but the finding the right number off beans and the range is often toe open toe interpretation. The box spot is clearer in how segments are defined. It's also more compact. You can feed several box plots in the space of a single instagram. 8. Time Series and Relationships: a second group of questions Do you will change over time on relationships. These questions focus mainly on groups off data points to identify patterns and trends, while a bar attacked will tell you what is happening on how entities compare Now change over time, we'll tell you how things involved. Making line charts helps discovering friends and cycling patterns, and they should be used only when there is some consistency in that change. If not about chart, you probably be a better choice. Another chart by William Playfair and again better than most line charts today. Note how he adds relevant contexts like references, two wars and no legend. He labels the Siri's directly and even adds an explanation. Here is a recent example I've made applying the same basic ideas, no legend context, the bailout period on annotations. There are a few issues with the regular line shark, but I'd like to discuss Onley one off them that display off time. Siri's with a strong seasonality. Imagine you have a long monthly Siri's. This chart displaying the data for the Arctic Sea extent is a good example. Other than the change at the bottom, it will be hard spot any change, so you can create a Siris for each year. Using this display, we are able to detect the overall pattern on surprise. Surprise. There is a winter, and there is a summer on. It's obvious that there is some level of variation, but it's hard to tell if there is the direction or just a random change from year to year. Having the longtime Siri's broken by year doesn't answer a simple question. Is the icy extent decreasing? Cycle ports are great to answer his question. In a cycle plot, you rearrange the time periods and instead off the regular flow, you compare all the appeals all the September's etcetera. This way you can detect changes in the overall pattern. Using cycle parts, the direction becomes apparent, and now we can answer our question. Yes, the sea ice extent is decreasing, especially in the summer months. While describing a variable is useful, we need to take the next step and see how they relate to each other. If, for example, we can tell that ice cream consumption is higher, it hotter temperatures, then we can plan accordingly. Visualizing relationships is one of the most relevant tasks in data visualization, and the scatter plot is the most commonly used chart type to let. Here is a scatter plot from Got Minor, showing that there is a strong linear relationship between income and health at the national level. The higher TV people happy toe the higher life expectancy when creating a scatter plot. The fundamental relationship must be defined by the variables in the X and A Y access, but you can call more variables into other properties. In this case, population size was encoded in tow, the dot size creating a bubble chart, and you can also use color drink of the categorical, variable like continent. These venerable shows that most African countries are to the bottom left of the chart, while most European countries are to the top right. So progression and continent actor nice to have variables that improve the analysis. But they should not distract us from the main insights. Finding strong correlations between variables is always interesting. It can also be dangerous because we are always tempted to assume cause ality. This scatter plot was made explicitly for fun on different Soto. Many people took it seriously. It says there is a strong correlation between chocolate consumption and the number off novel Laureates relative to population. It's easy to see that there are other variables at play here, but that's not always as obvious. So here are the takeaways for these two lessons. We can group questions burning on their focus. Some questions focus on the individual data, points like sort and ranking proportions and distributions. Other questions focus on groups of data points. There are other questions not discussed here in this class, including questions about connections, which belonged to network analysis on special patterns, which belong to maps and cartography. 9. Beyond the Single Chart: we focused too much on the single chart. We need to go beyond it and create visual narratives that glue everything together. You can call it graphical landscapes, storytelling or dash words. It depends on the context on the goals. It can start with the simple profiling and end with a freeform dashboard. There are types of questions that I like to call profiling questions. That means that you take a chart and replicated for multiple entities, for example, creating multiple population pyramids one for its region. You can see dozens of small and simplified versions of the same chart simultaneously, which allows you to detect different profiles. This technique has many names and sightly different concepts, but the most commonly used are small multiples. Trail is this place or scatter plot matrices. I believe that profiling should be understood as a single chart, not as just the post multiple charts because of its yes stalled. The insides you get are more or different than when you analyze the individual charts on the site milestones, industry off automatic photography, statistical graphics and data visualization. You can find these 1911 chart one of the first examples off this structure. We can use small multiples to compare entities or to show how an entity changed over time. In this case, we see how the retailer Wal Mart grew from a single store to a network that covers the entire continental U. S. But since this is a timeline, perhaps we could use animation and see how the chain spread like a virus. Animation is something that you should use carefully. In this example. Animation works because there is a simple and clear pattern. If that is not the case or there are details that you need to be aware off, small multiples will probably be a safer choice. This is one of the most compact ways off presenting data your is on chart. It combines the small multiples technique with access folding. Here, you can see how the monthly unemployment rate in each state of the U. S. The parts from the national average over the last 40 years. It's easy to see that some states managed to stay consistently below the average on DSO momentary spikes like this one in just by American Katrina. As the final type of question, we have the composition every form displayed. It uses multiple object types to communicate. Hear what matters is how the communication is built on how the visualization helps it to flow a good starting point toe. Have a consistent flow. It's bench neither Mons. Visual information seeking mantra overview. First zoom and filter, then details on demand. Here is an oversimplified an abstract example started the simple and perhaps playful chart , and once the reader is familiar with the core, message started telling it using more complex visuals. By the way, that's one of the reasons why banning part charts and other actually charts is itself not such a good that year. Remember how you should write titles for your charts? First, do it for its chart in your display, then read them out loud in the expected sequence and see if the message makes sense to my knowledge. The best book on this freeform visualization Steph Infuse book on Bash Board design. It is useful but naturally very specific toe dashboard design, so we do need more books on freeform visualization design. In this example, I used the Walmart later we saw earlier on adult population data to estimate population volume on profile in the catchment area. This other example, is a population. Dashboards are designed years ago, so the takeaways for this lesson, what is often a sequence off charts can be turned into a consistent message. Made off multiple visual paragraphs between the chart and this freeform display, there is a structure for the visualisation we call profiling. In most cases, we use profiling to compare entities, but we can also use it to display change over time with animation. Animation is useful, but only if the change is simple and clear without overtaxing our working memory. When creating freeform visualizations, we need to pay attention, toe how the message flows, just like written text. 10. Design for Effectiveness: much off the visualization part off the visualization is designed too much annoyance off graphic designers. We outsiders often see designers prettify ing things. It's not entirely wrong, but it's much more than that. Let's go back to the random PowerPoint chart exemplified these. We have these for data points on. Depending on what you want to say, you'll choose a line chart or a bar chart. These are strictly functional choices, but then you decide to spice up things a bit because the chart looks boring on you had through the decoration. There are many intermediate states between these two extremes, but the point is, you can make your charts more elegant without sacrificing functionality. In general, you should be able to justify each of your design choices, making them deliberate. You have to ask yourself, Well, this changes make the chart easier to read and understand. Will they grab attention without reducing function? Let's see how it works. This is a default chart. In an older version of Excel. I only added a smiling call. The data shows meat availability in the U. S. Over the last 100 years, I'll keep the original chart on the left and make some changes on the right. Since this is a professional setting, adding we part to your chart is probably not a good idea. So let's start by getting rid off the smiling CO. There are too many grid lines, and the Y axis doesn't need a decimal place. You can recognize this version of Excel by the useless great background, so let's remove it. Legends should be considered a necessary evil on removed whenever possible. Without legend, we can use the extra space to enlarge the plot area. Also, we should avoid vertical text by moving the Y Axis title to the top. Unless you're faxing the charter for some reason, need a black and white version. Markets are no longer needed. Color is a complex issue and should never be left to the application. The faults, if you have no graphic design skills, approach it from a functional perspective. Let's remove it For now, without markers, lines are almost invisible, so let's change that this is an important step. We are moving from a visual data dump to agree a visual message. We want to focus on the change in meat consumption, especially how, after many years people now eat more chicken than beef. Also, the chart is not the table. Andi, if you label every single data point, all we get is about chart under that table. Let's just label it relevant data point. Think of these as what the New York Times graphic desk calls the annotation wire, where you can add useful text orbiter notes to clarify or at context, we'll add more notes in the letter step. One way of making a chart or interesting is to use its title toe verbalize, major insights and conclusions. Here is a trick to do that. Complete the sentence As we can see in the chart. The text that follows will probably be a good title. So in this case, as we can see in the chart after declining since the late seventies, beef consumption was recently surpassed by chicken meat. Then you can use the subtitle to describe the date in the chart. Meet availability per capita in pounds in the U. S. On the new formatting detail. The border is useless here and can be removed. Find meaning Our message. The remaining meat sources are not relevant so they can be emitted a bit more park remains consistently high during the whole period, so we decided to identify it. Let's remove and minimize some more stuff. Grid lines are useful because they got the ice, but they should be muted. We know where the Y axis is so we can remove the line. Now we can label each Siri's directly. There are too many tick marks and labels in the X axis on. We don't need a four digit date that turning beef consumption is intriguing. Apparently it is explained by a price increase. So I added that no toe the annotation liar. And here is the final version in its full glory, a much simpler cleaner and pockets chart that is also either to say prettier than the original one. As you can see for each object and formatting option, there is a rational and functional reason. They impact aesthetics but making the chart prettier. Ease as we saw a by product, not a goal. It's interesting to compare the original chart of the chunk we removed from it. It's almost identical when using tools like Excel, especially in earlier versions. Much of a work consists of removing junk. The slight title is obviously a reference to one of my favorite data. Visualization offers Kaiser Fung on his blawg junk. Short in summary. Ensure that your design choices are functional and deliberate and focused on a message. Then find details that are cluttering your chart. Then use the extra room to add more data, added notations or simply have a cleaner look. Finally, used chart titles as newspapers, headlines and not a boring description that is more suited for a Met attack. 11. Functional Color: Golar needs to be discussed separately because it's really complex. In this point. I fully agree with Edward Lefty, who said that the goal is not to master color but to avoid catastrophe. Color is a fundamental part off aesthetics, but it's easier to deal with it if we approach it from a functional perspective. Let's see what I mean. If you look at this chart, you'll see that this series have a natural order 1st 2nd 3rd Now, if you look at Gore, including you noticed that there is no order is purely run them. So because the author failed to translate or the series in toward that color, including the chart is unnecessarily harder to read. Actually, we could avoid core, including altogether, with a different and more effective chart. So what I'm saying is that the chart fails not because colors are unpleasant, but because they failed at the functional level. Here is another example. Can you guess what this colors mean? Hope you came because I can't To me, that's just random colors, but we are left wondering if there is some hidden meaning that we can see. What did you change here? You need to ask yourself if there is anything to gain by using color variation. If the answer is no, then don't use it. If the author wanted to have a colorful chart, it would be much more informative. For example, color encoded a geography like continent. So in a moment I will commit a rosy off telling you that the actual color doesn't matter that much. But there is an exception. It's when you have to deal with its symbolic and cultural dimension, including brand colors. In those cases, it should respect the symbolic meaning because it avoids confusion and makes the chart easy to read. For example, if you're comparing meat vegetable consumption, you should use red for meat and green for vegetables are not the other way around. On the other hand, if you're comparing male and female, the blue and pink associations are easier to read. But if you want to avoid stereotypes, choose a more neutral pair of colors. But please don't teach them unless you want to make a point. And effectiveness is not your primary concern. So let's talk about the functional tasks off color. The first task is to categorize to color Inco distinct and independent categories. We can use car to group data pointing some meaningful way using a single car for each group or use variations off base you, By the way, if you are not using colors to group slices, there is no functional reason to use multiple colors in a pie chart. As you can see, instead of choosing colors, we should talk about chromatic stimuli intensity. We can use course to fuel arch areas because they require a lower chromatic intensity, while smaller areas need more saturated turns to make them easily identifiable. Playing with stimuli intensity also helps creating liars of relevance in a chart in this by I take to you to emphasize slice to in the focus plus context analysis, we often you score for focus on gray for context, as is simplified by packed bars, a chart type created bison Craig. We saw that if there is some kind of order in the data, the order should be preserved when using colors. Usually we do that by selecting a call around. This is a version of the population pyramid do so earlier, where each line is a year from 1985 toe 2050. Sorry, for the Samos plug. I used the diverging call a ramp in my book offer. I have a jingle. Ramps are useful to show differences to a central reference point in every good scale. For example, when it's is very from don't agree to agree, you can use car for a pop up effect when you want to alert users about something like a value below the limits off a variation band, no political or object among shades of grey is much more noticeable. That, among other card objects, this is useful not only to create alerts but also to create context and define levels off relevance. Similar letter. You Incontrera colorblind person who is unable to read your charts because of the colors who chose in these image. You can see a few regular color palettes on the left and assimilation off how they are perceived by a colorblind person. I used a small app called Color Oracle to exemplify this note, for example, that the color black person can't distinguish between the typical position off right on green. So that was the functional dimension, and the most relevant went for us. I would say that you will certainly avoid catastrophe. If you follow these ideas, the next step is to search for some harmony and you'll find it on the car will. What I find interesting is that the rules off your harmony played nicely with our functional tasks off color. For example, complimented records are in the opposite sides of the color wheel. You can use complementary colors to encode two distinct Siri's like male female. Our imports exports are you could use to groups off similar colors to compare European and non European countries. I'm sorry to say, but me it on designer functional, including, Is my primary concern not color harmony? This is how I'm able to justify my choices and feel that have much more control over the results. But that doesn't mean that we should start picking mostly random colors from the color wheel. A color pilot will ensure color consistency, and often it will be color blind safe, so you don't need to worry about that. You're too will probably let you select a color palette. Try to avoid a default one. If you don't like the pretty find ones, there are places where you can get pallets created by experts. Call a brewer is a very popular side. If you want a professionally designed palette, don't forget. Make sure that the palate you choose supports the functional tasks you need. So these are the takeaways. For this lesson. Make sure you need color. If you don't use it. Toe identify categories. You'll be able to use it to add relevant details. The data and the tasks often suggest the type of core, including use chromatic intensity to the note well of us. But avoid large splashes off saturated cars. Switch between multiple core pallets. Test which one feels right for your chart. 12. Resources: I focused discourse on data visualization for the office, but this is a diverse and expanding field that encourages cross pollination. So let's take a brother view. Let's start to the tools. Excel is obviously the sexist one. Then we have what we could call the visual statistics tools like jump or spot fire than the self service by tools like Tableau Barbie. I are quick view than the online tools, then the programming languages on Finally, Adobe Illustrator on its Adams like the recent Data Island Andi. If everything else fails, don't forget that you can use parent paper. Actually, I forgot to mention that pen and paper should always be your starting point, and you can also use play dough. This is a list of people that makes my forces syndrome skyrocket. I wanted it to be representative of multiple data, visualization, communities and profiles. Then you'll find many off these people in conferences like these. I advise you against following these links unless you are able to take a few days off. And finally, here are some books published recently. Check the Project on Resources Section, where I added links to these and other resources 13. Conclusions: office users are caught between bad excelled individualization practices on better but non applicable graphic design practices. I really of this can be easily changed with the right training and changing perspective. That was the purpose off this course. So desire some of its major takeaways always keep in mind that we visualized to get on communicating cites in the letter not to create aesthetically pleasing images, then exported data from multiple perspectives using absolute and relative values ratios and so on. Whenever possible, combined visual statistics with more traditional ones. For example, a chart could help you choose between average and median, or you create a scatter plot and add the D R squared. Be sure to know exactly what you want from the data and turned that into a precise questions. Now regarding the visual part approached the individualization as a kitchen robot or something that pre processes that that so that your cognitive resources can be allocated to more complex tasks. Because you have clear questions to be answered, you can focus on a design that is mostly functional. The design of the chart as a whole on the properties off each component should be cept on purpose and be justified. And finally, you don't forget a text part annotate Relevant data points have meaningful titles at comments that established bridges to existing knowledge. So where can you go from here? If you want to learn more about this idea off data visualization Safety net for Excel users My book goes deeper than this class. What you've learned here, essential to my approach to data visualization. My plan is to post other classes on a regular basis. They will be more practical and focus on the task or a tool, but they will share. We disc last the same principles and suggested guidelines. So thanks again for taking this class. Please review it on. Don't forget to upload your project. See you.