Friday, July 03, 2009

Junk Viz - Web Searches

Search Engine Land has a post, Michael Jackson’s Death: An Inside Look At How Google, Yahoo, & Bing Handled An Extraordinary Day In Search, on how web traffic spiked at some of the web's leading properties like Google, Facebook, and Wikipedia, as a result of Michael Jackson's death.

All good and fine, and a sad day for fans of Michael Jackson, the king of pop as he was known as, but a sad day also for data visualizations.
The chart above is a time-series area graph, and you can see that on the 25th of June 2009, around 14:00 hours traffic to Google querying "Michael Jackson" or combination of words thereof, began to spike. But by how much? Where is the scale? What does each increment of the gridline indicate? 1 million searches? 10 million searches? 100 searches?
Secondly, the area chart could instead have been replaced with a line graph, thus minimizing non-data pixels.

The bar chart above does a better job, in that you can actually see what the vertical scale represents. However, there are at least three problems with this chart:
  1. The color scheme makes it tough to see the data clearly. Of course there are only two bars, so it is not that difficult.
  2. The X-axis labels are gibberish. There is no sub-title or explanation of what these numbers mean. What does "6.4k" mean? And what do the zeros at the end signify?
  3. The location of the vertical scale on the right is non-standard. Most often a scale is placed on the right edge when there are two axes on the graph, as in a dual-Y bar/line graph, and the left and right edges both have different scales. For example, if you were plotting sales and units on the same chart, and using the left axis for the sales and the right axis for the units data.
  4. Adding a fourth quibble: time series data is best visualized by a line graph.


A better graph than the first one, but with the same problems of having no vertical scale.