When it comes time to present the process or insights of an analysis, visualizations are the best way to effectively and efficiently communicate that important information. While a lot of people are used to looking at tables of numbers, there are huge advantages in using charts and other visualizations. Most notably, it allows users to consume the same information, and a small fraction of the time that it would take to read it in a table. In one way or another, we have all been influenced by charts we have seen representing important statistics that are relevant to our daily lives. What is so fascinating though, is how much thought and design goes into what message the author wants to get across. By just changing a few aspects, such as axis scale, color, and title, the creator of the visualization can manipulate the information to paint different pictures. While it's great to know the great design can lead consumers to reach their creators intended conclusion, there are some visualization best practices, that can ensure data is presented in an accurate manner, as well as in interesting way. A few general do's and don'ts for creating data visualizations are as follows. Do, use an appropriate scale for your axes. All software that can create charts will automatically choose a scale for your data. Double check that the intervals being used are consistent. If it is displaying more than one variable, make sure the scale works for both. If one of the variables is skewing me access, it can make it much more difficult to accurately read the other information being displayed. In these instances, it is actually better to split the information up into multiple graphs, or use a filter to only show one variable at a time. Include means, medians, and trend lines. A chart can be much more valuable when you build in some benchmark, to compare the data to. Means, medians, and trend lines, can provide contacts that can help the users interpret exactly what the chart are saying. Knowing where your data lands in relation to what could be considered standard, and or how it is trending, are important insights that you will want to convey to your audience. Be considerate when using colors. When deciding what colors to use, there are a few facets to keep in mind. One thing that is becoming more common in practice, is shifting the mindset from the traditional, red means bad, green means good coloring, to either using blue and orange to show contrast, or switching to a single color gradient. Why are blue and orange recommended? Approximately one in 12 men, and one in 200 women, have a form of color vision deficiency or CVD. Blue and orange are two colors that are usually unaffected by those who experience CVD. Single color gradients are recommended, because they reduce noise in a visualization. Meaning users can quickly identify what stands out without having to translate what the colors mean. Finally, do a quick grayscale check. There are instances when your chart could be printed or viewed in grayscale. In some cases, the colors that are used end up looking exactly the same. Label everything. This might be self-explanatory, but it doesn't hurt to have a reminder. Clear and concise labels will help new users become more familiar with the content, and will make sure points are being read accurately. The don'ts. Don't pack too much information into one visualization. Simplicity is key to making impactful charts. I'm sure we've all seen line charts with 12 different lines showing 12 different variables. When you try to put too much information in a chart, the important information gets lost. For example, if you're making a line chart showing revenue of Grocers in a certain area, it is appropriate to have individual lines for the top three Grocers, and a fourth line that combines the rest into another category. If more detail is needed to show the revenues of the other category, it is best to create a whole new chart for that. Don't create 3D charts. Just because you can make 3D charts doesn't mean you should. Our minds are not good at recognizing depth or angles, making 3D charts very hard to comprehend. There are plenty of ways to make 2D charts, more interesting through creative design. So only use 3D charts when you are absolutely sure it's what is needed. Don't be too liberal with dual axis charts. There's a time and place for using two axes, such as when you are combining two types of charts. However, because we read from left to right, using two axes to show two variables, using the same type of chart can be difficult to read. Your eye is drawn to the axis on the left side. So at first glance, that is a scale you referred to, which is incredibly misleading. Now that you have an understanding of the best practices, there are some suggestions of best use cases for specific graphs. Bar charts and column charts. Bar and column charts are using all functions of an organization, because they are the best for comparison, and or explaining volume of different categories. These charts are easy to create, and when you sort the data to either ascend or descend, the user can understand the information in less than five seconds. Stacked charts. Overtime pie charts have fallen out of favor because it can be difficult for the human eye to determine angles. But a great alternative is a stacked chart. Stacked charts can either add up to the total or be set up to show the percentage of the total that each group contributes. Common charts that can be stacked are bar, column, and area charts. Which one you use will depend on what message needs to be conveyed. Line graphs. Line graphs are also used in basically all functions of an organization, because it effectively displays time-series data. These charts can show long-term trends, using actual data in an easy to consume format. Scatter plot. It is often recommended to make a scatter plot at the beginning of any analysis, because this chart can help the user see trends, but also identify any major outliers in a data set. If outlier detection is something that needs to be conveyed, a scatter plot is a great tool to easily show how many outliers are, and where they fall compared to the rest of the data set. Waterfall charts. Although waterfall charts are not super common in all functions, they are quite valuable for explaining financial trends. These charts are commonly used to explain profit and loss, because it provides a start point and an end point, and tells the user exactly what happened between the two. Data visualization is becoming increasingly important to making data-driven decisions, and creating impactful charts is really an art form. The main takeaway here, is to keep the user in mind, and create simple visuals that will clearly tell them the message you want or need them to hear. To take visualizations to the next level, consider combining them into a dashboard that will not just send a message, but also tells the story.
