It’s quite hard to define. Rob Austin at interworks.com suggests:
One would typically use this type of chart when dealing with data which has 3-12 different members in a few different dimensions. It is great for seeing where the overlaps are. When trying to to this type of analysis with other chart types you will either get an impression of the results without easy detail (e.g. stacked bar / heat map) or you will get the detail without a real overview (e.g. scatter plot). The Marimekko chart gives a really nice compromise in the middle.
No examples available at the time of writing (July 2016)
Rob Austin on interworks.com created this post back in 2013. It’s comprehensive and contains a five step guide to achieving the end product:
My introduction to Marimekko charts was through Andy Kriebel, whose “Data School Gym Challenge” proposed a visualisation in a different format to those shown by Rob Austin and Joe Mako. I actually prefer Andy’s method, but that’s a personal choice.
Jonathan Drummey contributed this in August 2016. It contains links to four blog posts outlining the end to end process, and it’s as complete a guide as you can ever hope to find:
Joe Mako supplied this example:
The example from Rob Austin is also available on Tableau Public:
Chris Love shared one in July 2016:
Andy’s Marimekko is here: