Be Faithful to Your Data – Avoid Improper Object Scaling

Data visualization is about using graphics or objects to represent the relationship of data, but some analysts are not faithful to the data. They manipulate the Y-axis (read my post here or one by Quartz here) or scale objects in ways that are misleading or just plain wrong.

You want to be unfaithful to your home team, your spouse, your diet and exercise routine – that’s on you. If I see an analyst distorting what the data has to say in a chart or dashboard, I am calling you out.

Object scaling issues typically arise when an analyst:

• Confuses height with area for objects comparing data

• Tilts objects to create depth perception

• Uses 3D objects

• Excludes labels

People are conditioned to perceive taller (or shorter) objects as proportionately more (or less) of whatever quantity is driving the chart. This makes perfect sense and is accurate for two dimensional objects like bars or columns that have the same width for every interval. Think about your classic column chart showing market size over time.

Problems can arise when using objects whose height and area do not change in a 1::1 ratio – think about circles, triangles, squares, three dimensional objects, and all kinds of pictographs. A classic example is when an analyst tries to show the market size, growth, and segment shares using two pie charts.

What typically happens when charting in Excel or PowerPoint is that the analyst can control the size of the pie independently of the values of the data. This can lead to unintentionally or intentionally mispresenting the data.

In this example, the 2015 pie chart representing a $3 billion market has height (diameter) of 2.6 and an area of 5.31 (in my PowerPoint deck). The 2018 pie representing a $5 billion market has a height (diameter) or 3.7 and an area of 10.24. So, while the market grew 66%, the area of the pies grew 92% – which equates to an extra $780 million in “market” as visually depicted.

Tilting and 3D - Style Over Substance?

Some analysts will tilt an object to give a sense of depth. I am not sure why. Maybe they think it looks cool or will stand out – or maybe they are trying to distort what the data is saying. The visual effect is to exaggerate the size (value) of the items in the foreground and reduce the size of the items in the background.

The pie chart on the left has been tilted and a wedge pulled out to highlight “Our Company”. If the analyst doesn’t include the data labels it is hard to tell exactly how “Our Company” is doing. It kind of looks like “Our Company” has a bigger share than ABC Co. and is maybe 1/3 the size of 123 Co.

The data distortion gets compounded with 3D chart in the middle because of the height vs. area issue and the depth perception issue. The initial perception might be that “Our Company” is 3X the size of ABC Co.

The graph on the right is the same data charted in a flattened pie chart with data labels. An entirely different story emerges with “Our Company” having the smallest market share – half that of ABC Co. and about 1/7 of 123 Co.

The best way to visually display this data (to show market size, growth, and share over time) is with a flattened, two-dimensional column chart. The objects (rectangles, wedges) are in proportion to the data. You can quickly see the growth and the share of the respective companies. Everyone is on the same page and can have a meaningful discussion and, hopefully, make a better decision.

The other area where you often see data distortion is in infographics and light marketing pieces using pictographs designed to catch your eye. The primary issue is confusing height for area to represent the data. The mistake the analyst makes (ok, it’s probably someone more interested in clicks than faithful representation of the data) is to increase the total size (area) of an object when trying to make the height X times the height of the first object to show the change.

The data says sales doubled, but the pictographs suggest they quadrupled.

For fluffy marketing pieces, I guess we can just snicker at the inaccurate representation knowing that we, as superior analysts, would never resort to such crude visualizations or intentionally misrepresent the data. But when it comes to creating charts and dashboards to support important business decisions, I will choose a boring flattened, two-dimensional rectangle over a sexy tilted, 3D object or pictograph every time.