Treemaps are useful for observing patterns among groups that have many levels. Treemaps are like bar charts that have been folded over in two dimensions so that there is no unused space. Rather than drawing a single bar for each measurement, a treemap can show the magnitude of a measurement by varying the size or color of a rectangular area. Use treemaps when your data contains many categories, to visualize many groups.
Treemaps were named and popularized by Ben Schneiderman, who has an extensive website about the idea ()