Stamen is a design and technology studio in San Francisco.

You are at Stamen's first blog, mostly written by Eric Rodenbeck. Visit our new blog! or come visit us at our newly designed

Contact Stamen, or follow us on Twitter

Recent blog posts

Sep 3, 2009

And the maps just keep on comin'

The State of California launched our most recent project today, a closer look at the $5.6 billion in Federal Stimulus funds that the State has received so far. There's alot there and I hope you'll go play with it yourself but the main thing I want to say is that you can slice and dice it all kinds of different ways and it's just slick as hell.

And this one is from a little while ago, but I didn't blog it when it came out—a live visualization of activity between computers for LogMeIn, whose recent successful IPO is leading some to speculate that perhaps all is not lost.

Both of these projects were headed up by Sha Hwang (whose bio I'm still waiting for) who's been doing some really great work at Stamen since he started about a year ago.

In local news, there's been a huge sting operation in San Francisco's Tenderloin aimed at reducing crime in one of the city's most troubled neighborhoods, which has apparently reduced things like property crime in the neighborhood dramatically. Since we've got access to the city's crime data for the dates of the sting, it should be interesting over the next few weeks to see how the map changes in response to this effort.

Criticisms of Stimulus Funds Map

We were just having a private discussion about the Stimulus Funds visualization, and in the interest of developing best practices for visualization and improving visualizations in general, I'm figured I'd post my comments here. Please take these comments as constructive criticism.

Note that I usually think Stamen does excellent work. However, I think the Stimulus Funds visualization fails both on the visualization side and the interaction side.

First, using point-based overlay visualizations like pie chart overlays to represent information about area-based geographic features is almost always bad design for two reasons:

  • Charts overlap for small areas spaced closely together. This makes reading the charts in an overview mode difficult to impossible.
  • Area fills allow a user to immediately see geographic trends, while charts require studying and comparing each chart to see the same type of trend.

In my opinion, geographic trending is such a critical aspect of geographic visualizations that it should only be discarded in rare circumstances (a good alternative for point-based data that keeps geographic trending would be density overlays).

Granted, point-based pie charts were probably chosen because you are also representing cities and wanted to be consistent, and doing multi-variate fills on areas is more difficult. Though rather than making counties into a point feature, it would have been more effective to make cities into an area feature (eg: using the zip code areas for each city as an approximation). There are some drawbacks with small cities with this approach, but in my opinion, fewer than with pie chart overlays.

Still, this map has made the classic point-based pie chart representation worse in several ways:

  • Placing the number of projects using a big white number over the chart, so it becomes even harder to read. This is unnecessary, since the size of the chart is determined by the number of projects. It would have been sufficient to let the size speak for itself and use a hover popup to display the specific number.

    [Note: I just realized the size might represent the total budget; this is unclear. Nonetheless, with the correlation between number of projects and total budget so high, it would have been better for a popup here].

  • Failing to pop the focus chart to the front when the mouse hovers over it. This makes it so you can never see certain charts in the overview mode.

  • Failing to draw county boundaries. While this makes for a cleaner design, it loses a critical aspect of the analysis: the size of the county. So no comparing number of projects to county size. It also makes it difficult to know where one county ends and the next begins.

  • Using a bright pure white thick border, which adds chart junk to the visualization. While it's important to separate the charts, especially since they overlap; a one-pixel border with a dark color which doesn't grab the eye as quickly would be better.

  • In the zoom in mode, county information is only shown in the center of the county where the point-based pie chart is; if you pan around the county, the information disappears as soon as the pie chart does.

While I like the sidebar, it too has problems. It's unclear what the colors represent (I thought it was the largest number of projects of that type, but that's doesn't appear to be the case since LA's pie chart shows Energy as the most, but the sidebar color is for Transportation). Also, the hover bar that tells you information about the focus on the sidebar doesn't appear when only the county name is visible (thus, you have no idea how many projects or how much was spent for that county).

Finally, the visualization lacks a good legend or popup help describing what the various visual attributes mean (ie: size of pie charts, colors on sidebar).

Overall, I think it would have been more effective to represent a single metric at a time using an area fill with a control that let's the user switch between the total of all metrics or the slice of a specific metric. This would more clearly show which counties and geographic areas of California were getting which types of stimulus mone and allow you to compare size of county to the amount received. Two variations of this would be to use small multiples to create one map for each category (since there are only 10 categories), or to use a multi-variate fill on the areas (eg: categorized colors like a pie chart, but used to fill the area, either in a circular fashion like a pie chart or in a stacked fashion like a stacked bar chart).

Post new comment

The content of this field is kept private and will not be shown publicly.
More information about formatting options