Wednesday, January 14, 2015

Jon and I chat about non-zero baselines

Earlier this week, I got a rare chance to chat in person with my friend Jon Schwabish ( who was in town for a quick trip. We recorded a short chat on a topic that's been generating some buzz in the data viz world lately: non-zero baselines (which I've previously blogged about here and here) and whether it is ever ok to start the y-axis of your graph at a value other than zero.

Check out the following to hear our chat. Leave a comment here or on Twitter (me | Jon) to join in the conversation!

A couple visuals we referenced while talking:

The graphs we had up on Jon's computer as we were chatting
From Drew Skau's recent blog post

1 comment:

  1. Thanks for an interesting discussion – I do subscribe to the position that context is critical when presenting any graphic visualization. However, my issue is with the first sample graph that you have used, and how it illustrates the issue of lack of context!

    Based on the X-axis values, if we assume (!) that the first three data points represent decade-long periods, and the fourth data point seems to only represent a 4/5 year-long period, then the data points represent different aggregations, and should not appear on the same graph. The visual drop in value is actually misleading. Any other interpretation to justify the sharp drop would need a defined context and explanation of the X-axis values.

    An alternate approach to the issue of non-zero baseline graphs is to consider representing the data as “percentage change” values, rather than actual data values. This would allow the graph to show, using a zero baseline, where a 0.1% to 0.3% change is relevant, or even a -5% to +10% is what is important.