[MUSIC] Hello and welcome back, I'm Halden Williams. In this video, I'm going to review some academic theory that can help us better understand how to develop effective data visualizations in Excel or across other tool sets. As we've previously discussed, effective data visualization helps an audience analyze, understand, and draw conclusions from summarized data. Building on the principles from our previous lesson, let's add the following recommendations for charting. Less is more, your chart should convey one message to your audience. Communicating multiple concepts usually results in tradeoffs in clarity. Use titles and labels to give your audience an understanding of your message. You want your audience spending their time exploring and thinking about your message and the underlying data, not orienting themselves to what they're looking at. Use call-outs to highlight key messages and direct the audience to findings. Use color intentionally and not for decorative purposes. Less color is often better, this is consistent with a fundamental principle we discussed in an earlier video. Understand constraints of your medium, how will this graph be presented or printed. If our visualization is printed in black and white will that change the message? One of the academic pioneers in data visualization coined the term chart junk to refer to any nonessential information added to a chart. A good test for your chart's effectiveness is to try and remove an element from your chart and assess if your audience will still understand your message. If they can, you should remove it. Let's see if we can improve the effectiveness of the example chart by applying some of our new concepts. Before we start, let's review the message we're trying to convey. This chart should communicate that Chicago is the third most populous city in the country in 2016. On the screen we have a bar chart for the population of major US cities, in its current form, the chart makes it much more difficult than it should be to answer this question. Let's see what we can do to make this graph more intuitive to the reader. To start, let's make all the bars the same color, applying the concept of similarity. There is no immediate benefit to have each bar a distinct color so changing them all to the same color makes it easier for us to look at. Next, let's examine the Y axis and data labels. Do we need both? Probably not, furthermore, is there a benefit in having the exact population for each bar? Or would it be to use some rounding here? Let's try rounding to the nearest 100,000 for data labels and remove the Y axis label. This looks better, however we've removed the scale and added rounded values. So we need to ensure our audience is aware of the scale. We'll revisit this when we update the title. Now that we have data values and no Y axis, let's remove the grid lines as they no longer provide any benefit in this example. If your chart does require grid lines we recommend using a light gray instead of black as a grid line should not be the focal point of the chart, rather a guide to helping the reader interpret the results. This charts looking better already, let's see what else we can do. The order of the bars is currently alphabetical by city. Let's change the order so the cities are organized by population instead of alphabetically, as well as flipping our chart orientation so it's easier to read. Great, now that we've updated the order of the cities, let's review the title of the chart. Population's not a complete title and doesn't help the audience understand what are looking at. We can be more specific, the title of the chart should convey the primary conclusion succinctly. Let's update the title to read, Chicago is the third largest US city, 2016, (population in millions). There are few more elements we can eliminate and change around before this graph is finalized. Let's get rid of the border as it's not contributing to the story. The final touch will be to highlight the focus of our message. Since we are focusing on Chicago, let's use the anomaly principle and make that bar a different color. And there you have it, a much more effective graph. Let's compare this to where we started. You can see how we put some of the charting concepts to work. Telling a better story and removing non-essential chart elements. Similar principles can be applied to tables, as they're just a different type of visualization. Many times, a table can more elegantly convey the primary components of a data story without the added distraction of needless visual elements. Tables need to be a legible as possible, some characteristics of a legible table are displayed in this comparison. Banded rows or columns allow for easier distinction when matching and reading across information. Cell padding or adding a slight margin and evenly distributing columns when appropriate creates more space for the reader, and enhances readability. Creation of table should follow similar fundamental principles applied when you're creating charts. Data in tables should be organized in a logical order based on the message you're trying to convey. And where appropriate round to allow for easier comparison. Now that we have seen effective charts and tables, let's review some classic charting mistakes. In the first example, using a pie chart when the data points are very similar. It is difficult to draw a conclusion from a pie chart with many slices as we cannot easily understand the variability. An example of chart junk that should be eliminated are charts with a third dimension that communicate no information. Another problem with a third dimension is that it can skew the perception of viewer, which should not be your intention. Last and perhaps most importantly, often times pie charts do not convey conclusions which cannot be effectively drawn from a table. Further to the topic of skewing data presentation inaccurate scaling is another common theme in ineffective charting. It is important to remain consistent while you're using a scale. In this example, we can see how the difference being called out is grossly exaggerated with the inaccurate scale on the left. These two graphs show the same bars, but clearly biased representations of the data. This is often referred to as the lie factor. Infographics can also be guilty of the same inaccurate scaling, differences in width and height that are often results in visuals such as this one are misleading to the reader. While infographics are more interesting to look at it's important to consider how the message is coming across. One final common mistake in ineffective charting is poor, missing labels while it's unlikely you would create a chart with no labels at all. It is common to see a title or axis label missing. Without this information the the audience is forced to make assumptions, some of which maybe correct, some which maybe not. Similarly ineffective is including labels which are so disconnected from the visualization itself that they require multiple interpretation steps to understand. Take the guessing out of it and use all necessary labels, placed ideally as close to the value they are associated to as possible. This will also help ensure the reader is getting the message you are trying to convey without having to exert undue effort. The next videos in the course will move on to the details of how to create some of these charts, applying the principles discussed, and rules of effective and ineffective charting. During the remainder of this week, the courses will go through how to build visualizations in Excel. As we do this, keep in mind the more advanced environments discussed earlier will not be covered in this course. Excel provides a lot of flexibility to practice building visualization and serves as a foundation for more advanced data visualizations to come. [MUSIC]
