An unbiased approach to data visualizationAn Unbiased Approach to Data Visualization

As executive council members, CapTech attended an intellectually stimulating event at Georgia Tech's Business Analytics Center.

Keynote speaker Andrew Wells' topic was monetizing your data. Data monetization strategy is a plan to achieve one or more business goals' quantifiable benefit. It focuses on an organizational approach, tools, techniques, data and application. This blog is not a deep dive. The discussion prompted me to revisit how I approach data and visualization.

Even though we are racing towards machine learning, artificial intelligence and natural language processing, human factors affect the way data is interpreted. We can intentionally or accidentally apply biases to facts and dimensions. Displaying data accurately is not our only responsibility.

We should be mindful of biases when developing dashboards.

Diminishing sensitivity effect is the focus on relative differences instead of actual difference. In visualizations, loss looks bigger on smaller numbers. Business Intelligence (BI) developers should be deliberate in displaying axis and labels. In my charts below, the maximum number looks dramatic. However, the value is only 6.

Exhibit 1

Labels can clutter graphs, but they provide perspective.

labels provide perspective

Exhibit 2

Another example of relative differences (hate the pie chart, love the donut). When displaying ratios or percentages, it is helpful to show underlying numbers to provide context. Make it obvious that 100% of 1 is much different than 69% of 1,033.

This visual does not show the entire truth Numerator and denominator adds perspective

Numerator and denominator show perspective

Recency bias places emphasis on the last data point. In my example below, February looks like it took a deep loss. However, it is only mid-February. In this case, to avoid the perceived data plunge, the line should stop when the data is not populated.

Exhibit 3


Ikea Effect places higher value on something you built. A benefit to the Ikea effect is that you can get buy-in from users by engaging them in the building process. Feedback enables users to contribute to and shape the solution. If users are a part of the build, they will tend to accept and adapt. Involvement can be important to the success of the roll-out.

Ikea effect and user feedback leads us to another data monetization theme, question driven development. It focuses on asking questions to understand decisions. Why are we building this solution? What questions will the data answer? Take time to understand primary business questions and drivers, instead of racing towards a solution. Questions shape the tools, approach, requirements and interaction with the data.

In closing, I don't believe there is a magic formula to visualizing data. It is an art and a skill that requires listening and deliberate design. Develop pretty and interactive dashboards! Show the entire story. Do not lose sight of business questions and the users empowered by the data. Be mindful of biases when displaying data.