Tableau graph showing Gartner’s customer survey results

Very disappointed in Tableau for the production of this graph on their most recent blog post. I’m a big fan of their software and their credible stance on principles relating to data visualisation but, on this occasion, they’ve produced something which breaks a number of their championed best practices.

The graph displayed relates to the results of a Gartner customer survey and presents how customers are using various BI platforms.

The accompanying narrative is as follows:

In the graph above, we’re seeing the percent of customers using the product for one of eight different BI activities. Each stacked bar represents the summation of those. Since one customer can use a product for multiple functions, the stacked bar reaches well over 100%.

It would appear that for each of the eight categories of BI activity respondents gave a yes or no reply depending on whether they use the software for that purpose or not. So for each activity the potential highest value would be 100% usage.

Tableau are therefore trying to show how versatile their software is by revealing the high usage percentages against the categories as an aggreate, however, this is not an effective way to present the information. The value axis labelled ‘percentage of respondents’ goes way beyond 100% reaching about 260% which typifies the confusion the reader has trying to interpret the graph.

I haven’t got the raw numbers to usefully create a redesign but a more effective way to present this (and incidentally a method which is fundamental to the functionality of Tableau) would be through small multiples, perhaps with Tableau’s values highlighted. A summary graph could then be plotted revealing the mean percentages across all eight categories which would reveal tableau as having the highest value, thus achieving the apparent aim of this graph.

5 Comments

Elissa FinkMarch 29th, 2010 at 6:44 pm

Hi Ben,

Thanks for noticing our blog post. The graph was not created by Tableau – it’s Gartner’s graph.

I should have made it more obvious that the graph was not ours (and because of your comment, I’ll make that change on the post). Gartner does not release the raw data and only allows people to show their original graphics. So all we could do was show their graph.

I should have cited them as the maker of the graph; we know that there are better ways to show that data.

Thanks again for noticing our post!

Best,
Elissa Fink
VP Marketing
Tableau Software

Andy KirkMarch 29th, 2010 at 6:57 pm

Hi Elissa
Many thanks for taking the time to post this correction – most appreciated and also most reassuring as well!
I launched this site last month and so don’t (yet!) command a particularly large readership, however, I will post a follow up to ensure your comment and its explanation reaches the same people who might have read the original item.
Regards
Andy Kirk

Elissa FinkMarch 29th, 2010 at 6:57 pm

Hi Andy,

I’m embarrassed. I wrote “Hi Ben” when that’s not right. Please accept my apologies!

Best,
Elissa

Andy KirkMarch 29th, 2010 at 7:00 pm

Not at all, I’ve had worse!

Jon PeltierMarch 31st, 2010 at 11:57 am

Andy -

I actually took the time to “manually digitize” the chart, and ran through an analysis of chart types, culminating in a panel (small multiples) chart in Stacked Bar Chart Alternatives.

Elissa -

When I couldn’t activate a Tableau Public exercise by clicking on the chart, I assumed it wasn’t your chart. I know anyone at Tableau could make a better representation of the data.