January 13, 2010

Making Information Visualization work

Information Visualization is all about helping users develop insight. It isn't enough to show beautiful and amazing pictures - even interactive ones. The key, as with all interfaces, is to think about specifically what tasks a user will be able to perform with the tool. Yet, all too many visualizations fail on this most important and fundamental test.

Stephen Few (of Perceptual Edge) wrote a beautiful 10 page report on this issue - critically analyzing a visually simple visualization by Ben Fry of healthcare spending (see pie chart above). The key is that it doesn't actually enable you to learn anything. He then goes on to create an interface that actually does enable you to discover all sorts of things. The new interface is not simple. Well, each component is simple, but there are lots of components and it may take a minute to figure them out. But no more than that, and then you discover that it is conceptually simple, and you can actually use it.

Here are the take-home lessons from the article - but go and read it - it is well worth it.
  1. 2-D position and the lengths of simple objects such as bars encode quantitative values in ways that are easy to perceive; angles and areas do not, and therefore should be used only when you can’t use better means.
  2. We cannot build a picture in our heads of a pattern that is formed by multiple values (such as the average cost of healthcare for patients of each age from 1 to 79 years old) by looking at one value at a time.
  3. Lines do a good job of showing the pattern formed by a set of values across a continuous range such as patients’ ages, and do so in a way that allows us to compare patterns when multiple data sets are represented at once (such as one line per disease).
  4. Multiple graphs shown together are often a better solution than a single graph, especially when several variables are involved.
  5. Several simultaneous views of the same data set, each showing the data from a different perspective, make it possible to see relationships that can’t be seen from one perspective only or from viewing different perspectives independently.
  6. The ability to easily fi lter out data that doesn’t concern


No comments: