Radar’d to death

Having picked up a recent blog post by Stephen Few about a graph that had been promoted as an example of good practice by Oracle’s BI suite, I was searching around for reports on BI vendors and came across a decision matrix report by DataMonitor from 2007. The report I unearthed was on the SAS site, presumably because SAS came out very well in the recommendations.

My reason for sharing this is simply because it has ironic connotations similar to another report I came across in that it attempts to make a credible assessment of the BI industry but entirely undermines this (in my eyes at least) through its ineffective use of graphs through the report.

This first graph to comment on has one key flaw – the undermining of its integrity caused by the axes labels not starting at zero (even though the label says scale 0-10) which creates a skewed view of the plotted proximity of the various BI vendors.

The next graph type (of which there are three) is simply a shocker. It is a multi-variate radar graph plotting assessments of the 10 vendors across 12 different characteristics. Once again the scales don’t begin from zero. The more effective alternative would have been to plot small multiples of bar-charts for each vendor displaying their scores against each characteristic. Similar treatment was provided to this graph by Jon Peltier.

The final set of graph types – there are 12 of these – once again plots measurements of various activities, against various categories, referring to various types… I would make a greater attempt to accurately describe the purpose of these but, as I say, I got Radar’d to death.