Oftenly, business analytics is about telling a story, a convincing story. As they say, a picture is worth 1,000 words, and numbers alone, are a poor way to tell a story. Millions of years of human evolution have attuned us to respond to movement, direction, size, shape, and colors. Not to columns and rows of numbers, unless perhaps you come from a long line of bookkeepers. So let's not fight our animal instincts, but rather appeal to them with things like meaningful imagery, evocative of herds of animals or attacking predators, slinging spears, moving clouds, rushing waters, mountains and valleys, falling rocks, growing foliage, landmarks, and the changing seasons. Selecting the right type of analytic visualization or chart seems easy enough. But once you realize the near infinite variety of illustrative possibilities, it can become a bit overwhelming. Even choosing among a common pie chart versus a bar chart or a line graph, can leave you scratching your head. So let's look at a variety of common types of visualizations. First, let's look at tables. Tables typically display information in columns and rows. Think of a spreadsheet or a financial statement, as such tables really only lend themselves to two-dimensions of data. Sure you can convey a lot of information with a table. They can have an unlimited number of rows and columns, and tables can scroll up and down, and left and right, but they can be difficult and time-consuming to interpret, and or not at all intuitive to our primate brains. Tables especially spreadsheets are often interactive, allowing someone to sort, search, and filter them. In this way, they're great for data exploration but not necessarily for storytelling. Tables are also good for combining numerical and textual information, typically in columns alongside one another like student name and GPA, or product price and product description. Advanced forms of tables will use visual indicators like bold text or colors or flashing cells to indicate the more important data within them. As always, be careful when using colors to discriminate data as up to seven percent of the population is colorblind to some degree. Then there are bar charts. Bar charts are good for showing the magnitude of instances of something using the length or height of a bar, like the scoring average for basketball players on a team. It's not just the magnitude that's important, but bar charts can illustrate the relative magnitude of each. Bar charts are limited by their ability to show more than a handful of instances before they get too crowded. For example, a basketball team with 15 players may be possible to illustrate, but a baseball team with 25 players or an American football team with 55 players might make a chart too crowded and make the labels for each instance impossibly small. As such, bar charts lend themselves better to showing values associated with categories of things rather than each individual instance. Common variations on a bar chart can include stacked bars in which different values totally 100 percent or some standard maximum are placed one on top and other often using different colors or shading, or you can group instances of bars that correspond to a particular category. The biggest challenge with bar charts, is handling many instances or categories, finding a room on them for long labels, and that they take up a lot of space much of which may be unused or blank space. Then there are pie charts. Pie charts are ideal for showing relative portions of a whole in a way that's highly visually intuitive. However, once a pie chart has too many slices more than half a dozen or so, users will leave, well, feeling hungry. Pie charts also take up a lot of real estate on the screen and are limited to just one set of categories and one set of values. Perhaps, their beauty and their ubiquity are a result of their simplicity. In a pie chart if you want to emphasize one or more particular segment, you can explode them out from the chart. If you're still hungry, doughnut charts or like pie charts with no center. They are meant to show proportions better than simple pie charts. Other variations on a pie chart include polar charts where slices all have the same angle, but some are larger or protrude from the center more than others. Ring charts, sunburst charts, and multilevel pie charts. A line chart or a line graph, often illustrates data as a series of connected points. They're ideal for showing measurements over a period of time. Some line graphs rather than simply connecting the dots directly, will use statistical methods, again like least squares to fit a curve to some linear equation or to do a more complex model like an exponential curve or a quadratic formula. Curve fitting can help identify causal relationships and extrapolate future values for predictive purposes. Some line charts will show the relative area beneath or between each line to convey the total magnitude of a variable. These are sometimes referred to as steam graphs. They are clever, but really only useful when a handful of lines are included. Scatter plots are ideal for identifying patterns, correlations, and clusters among two different measurements, especially when the number of measurements is great enough for visual patterns to emerge. As such, scatter plots are limited to two variables only and may not be effective if the number of measurements crowds the dots too close together or overlaps them. Bubble charts, solve the two variable limitation of scatter plots. In scatter plots, all the dots are the same size. Bubble charts allow a third variable to be represented by the size or area of the dot or marker. The size of the bubble is usually used to show the magnitude of something. Bubble charts can also use color to add a fourth variable, but this can make them difficult to interpret. Bubble charts are sometimes animated to add a time dimension enabling user to see the growth or movement of bubbles over a time period. Unlike scatter plots, bubble charts are more limited in the number of values they're able to display without too much overlapping of bubbles. There also not that effective of showing trends or extrapolating on data for predictive purposes. Once a bubble chart has more than a few dozen bubbles, their efficacy in telling a story or conveying insights can quickly diminish. Then there are gauges, meters, and bullet charts. Gauges and meters are like dials on the dashboard of your car or markers on a measuring tape. They offer a familiar way to illustrate data of a single variable. They are also effective at indicating ranges or thresholds like high medium or low or safe, unsafe, and dangerous. Typically, gauges and meters are used as part of an overall analytic dashboard, again like on a car or as part of an overall infographic which I'll discuss momentarily. Bullet charts are simple line segments, superimposed on a set of threshold segments often colored. The line segment illustrates a range of magnitude within a designated minimum and maximum. For example, a bullet chart can display the actual high and low temperatures of a refrigeration unit during the week. A longer line segment may indicate that the refrigeration unit isn't as responsive as it should be, while a short line segment may indicate that the compressor is turning on and off too frequently and may lead to quicker part failure. Then there are heatmaps and tree maps. Heatmaps and tree maps often look like abstract art. They are matrices, they use different colors or shading and area sizes to illustrate the relative magnitude of different categories. With heatmaps, each cell in the matrix is the same size like a checkerboard, but are colored differently depending on which values are considered hot or important urgent or prevalent. As you can guess, usually the red cells are hot and the blue ones are cool. Heatmaps usually use a gradient of colors. With treemaps, which to me don't really actually look like trees, there are no x and y-coordinates. Rather the area of each segment is used to represent the relative magnitude of some variable. Sometimes you want to show a line chart or a selection of a line chart that illustrate relative trends for different items like stock indices, but you don't have the room or need to include all of the axes, labels, and coordinates. In this case, you would use sparklines. Think of them like mini line charts. Sparklines are great for including in-line in text or in documents without taking up too much room, but they're not good if you need to express actual values. When several sparklines are included on a chart, this is referred to as a small multiple, basically a multiple selection of small charts. Then there are spider or radar charts. No surprise, spider charts look like the analytics of spider would produce. Spider charts, again also called radar charts, are somewhat like line charts that go into a bit of a circle. Each spoke on a spider chart is of an equal angle and can illustrate the magnitude of multiple variables. Many more than a dozens of variables, an a spider chart quickly becomes unwieldy. Spider charts can be particularly useful when variables are related, but not to show any particular chronology or time axis. Network diagrams show relationships represented as lines, between items represented as nodes. They don't have any particular form. Sometimes the length of the line can be used to represent some type of proximity or similarity or strength of relationship between the nodes. Network diagrams are good for illustrating clusters of information or the affinity among items. They can also be three-dimensional with software that allows a user to travel through it exploring values and relationships, much like a starship traveling among galaxies, solar systems, and planets. Then there are maps or cartograms. For any geographic information, map visuals are indispensable. Sometimes data can be overlaid on a map, for example the population of different countries. Other times, geographic areas can be enlarged or be shrunken to show a difference of magnitude. For storytelling and often for promotional purposes, infographics have become quite popular. An infographic displays an artistic or aesthetic arrangement of numbers, charts, and text. Infographics tend to be quite large, requiring scrolling or zooming to see the entire picture. As such, they make for great wall posters. There are numerous other types of charts like subway maps, box plots, Pareto charts, Gantt charts, organizational charts, flow charts, pedigree charts, waterfall charts, and candlestick charts. As you become more adept at visualization, you may choose to use some of these. Whichever visualization you use, you may want to follow some general guidelines. Usually, you want to illustrate the most important and often used items at the top-left and larger. That's where most humans first look when reading a page or looking at a screen. Accordingly, you want to play supplemental and lesser used items at the bottom and to the right hand side. Normally, you want to stick to a few types of analytic representations and remain consistent in doing so. Mixing and matching pie charts and bar charts for example, can really be confusing. Infographics and dashboards are an exception in which users may expect a variety of interesting visualizations. Of course, you want to avoid crowding the canvas. Take liberal advantage of using whitespace. Next, you want to do away with distracting dashboard detritus like excessively long labels, descriptions or text. Interactive visualization should make liberal use of hover capabilities to display text. You want to avoid mixing and matching too many colors, shading, and fonts. But again, remember that 7-10 percent of men have some degree of colorblindness. So multiple concurrent visual indicators may be warranted. You also want to be sure to use consistent measures and to convert them automatically to international monetary and metric measurements based on the user. Finally, remember that your visualization will likely be viewed on devices other than the oversize pixel laden monitor you're designing them on. So ensure that they can be viewed, down-sampled or converted on any device, be it a PC, a tablet, mobile device, and maybe even heads-up displays invoice, for hands-free or visually impaired individuals. In many government departments, such accommodations are mandated by law.
