I came up with this term last week, during a presentation by Bill Lay at the Masters Summit for QlikView in London.
Low-cardinality part-to-whole comparison
For me, it perfectly describes the correct use of a pie chart. Let's break it down:
In database parlance, this means that a column has few unique values. For example, a flag field - either 1 or 0 - has only 2 unique values. An opposite example, high-cardinality, might be Account Number - lots of unique values.
In a pie chart, for me, this means that you need to have a low number of segments. Really only two or three. Too many segments makes it hard to discern the differences between them.
This is critical for the correct use of a pie chart. A pie chart is all about ratio - while you are comparing a segment against other segments, the correct context is the size of the segments versus the whole - what is the ratio of one segment versus the others and against the whole.
As an example, if we are looking at sales by country and look at only 3 countries - say Germany, USA and France. In a pie chart we might see that Germany has about half of the sales. But the context is just the sales of 3 countries and you can come to the incorrect conclusion that Germany is responsible for half of sales. Showing all of the sales for all countries - probably restricting to Germany and Others or Germany, USA, France and Others - then we see the whole picture.
If we just want to compare the size of one country's sales versus other countries, then a bar chart is ideal. The bar chart also works if you can't see all of the countries. If you only present a sub-set of the countries in a pie chart, the context is incorrect.
So, there you go - low-cardinality part-to-whole comparison. Feel free to use it in your next presentation.
Stephen Redmond is author of QlikView for Developer's Cookbook
He is CTO of CapricornVentis a
QlikView Elite Partner. We are always looking for the right people to join our team.
Follow me on Twitter: @stephencredmond