6 Tips to Make Data Visualization More Effective
Simple tips everyone needs to know to get information across as clearly as possible
Data is more widely available and useful than ever before. According to the Bureau of Labor Statistics, data scientists are among the fastest growing professions in the US. Of course, with the increasing relevance of analysis, it’s more important than ever to communicate information effectively, and data visualization is one of the most powerful ways to do so.
While data visualization is common, it’s not always used to its fullest potential. So here is 6 basic tips to make your visualizations as effective and easy to interpret as possible.
Use A Trendline To Make Patterns Clear
Patterns aren’t always obvious to the naked eye, so adding a trendline is a good way to clearly show the pattern while keeping the raw data visible the reader.
However, you should also be sure to pick a trendline that fits the data relatively well. The chart shown above uses a linear trendline, but you can also use other trendlines (such as a logarithmic or exponential decay) if the data displays a different pattern.
If you’re unsure, play around with a couple different fits to see what looks like it tracks the data best.
Smooth A Volatile Time Series
Sometimes a time series exhibits a trend, but also has a great deal of volatility. This can make a typical line chart painful to read (not to mention not aesthetically painful). Luckily, this can easily be solved by adding a smoothing method while keeping the individual data points on the plot. Not only does the chart become more aesthetically pleasing, this makes the trend easy to see and prevents the reader from focusing too much on outliers.
Above, I use loess smoothing but you can just as easily use a moving average or exponential smoothing if your preferred program doesn’t easily allow for loess smoothing. Smoothing methods are also very flexible to make them more or less responsive to outliers and large changes in the data. A real life example of this can be seen in how NASA displays temperature data:
A Pie Chart Isn’t Always The Wrong Choice, But It Is Most Of The Time
How the mighty have fallen. Pie charts are now widely derided in the data analytics community, and for (mostly) good reasons. In almost all situations, there is a better way to display the data than a pie chart.
So what should you use instead? If data from previous years is valuable as well, consider a stacked bar chart or even a stacked area chart that might show the trend over time. A simple line chart might also do the trick. If you’re only interested in current-year data, then you might also consider a bar chart. If that isn’t a great fit for your data for whatever reason, then you might use a pie chart as a last resort. One common situation where pie charts are still used is to display market share among a group of competitors when focusing on the present-year.
Annotate To Draw Attention When Needed
Sometimes you might want to draw attention to some specific data point or add additional information to your chart. To do this, you can add text annotations to your plot to expand on information. Here is an example of a chart I made some time ago:
In this example, I use text annotations to label data points that are interesting (the outliers) and to add additional information to the plot that would otherwise be impossible to discern (that 25% of movies lose money).
However, text annotations aren’t always useful. If there is nothing that requires annotation, don’t feel a need to force it onto the chart. You should also always keep in mind that too much annotation can make chart feel cluttered and reduces readability. Use your best judgement to decide when and how much annotation is needed.
Add Labels Directly On The Plot
Traditionally, legends were added to the top or side of the plot. However, your plot can be much easier to read if you add the labels directly onto the plot. Make sure to match the color of the line to its corresponding label when possible to make the effect obvious.
Not only does it look better, it increases increases the amount of space your plot can take up as well, making it easier to read. An alternative is to place the legend within the plot where there is a good deal of whitespace — in the example above, the top left is a perfect spot.
Of course, sometimes this might be difficult given your plot type, but whenever possible this is usually easier to read.
Choose A Good Y-Axis Starting Point
To make a compelling visual, you need to choose an appropriate y-axis starting and ending point. Some people will advocate that all graphs should start at zero, but this approach is far too strict and can destroy the readability of many charts.
Unfortunately, there is no great rule-of-thumb for where to start your y-axis. In many cases, starting your y-axis at zero can make the graph difficult to read. However, zooming in can make small changes in the data seem much larger than they are. You need to consider the trade-offs in every unique situation to decide if starting your y-axis at a value of zero or something else is best.
