I went to my first SigChi meeting last night at the headquarters of Attachmate in Bellevue, WA. A dork among dorks; I arrived an hour early. First to fill out my name label, I wrote...

Korby Parnell

...in big bold letters. Apparently, I am the only geek on the Eastside besides my friend Jonathan that has been blessed with both a last name and an employer whose name I can wear with pride. "Hi...Mike... I am Korby Parnell from Microsoft." Okay then, moving on. The mini-sandwhiches were fair, the spinach dip watery, and the vegetables unblanched. But conversation was lively (Hi, Rebecca. Hi Jalli.) and the speaker, Jock Mackinlay had an interesting presentation. However, I was more inspired but what he did not say than by what he did. More on that below.

[Rajat Paharia] "Although Pat Hanrahan and Jock Mackinlay are obviously smart guys... Mackinlay worked in the User Interface Research group at Xerox PARC, which came up with a bunch of very cool data visualization tools, some of which were commercialized by Inxight. One of the things that most of the Slashdot posters seem to have missed is that it's not just a pretty graphing tool, but you can manipulate the graphics - which in turn manipulates the underlying data. Now that's pretty damn cool."

To build on that, you can dynamically select a region of interest in a Tableau Workbook graph, drag and drop it to a shelf for later inspection, or export it to another application. You can even drill down into the data underlying its visual representation. Unfortunately, I do not think that Tableau enables you to do Updates in addition to Selects.

For example, imagine that you're a test lead. Your team uses a SQL Server-based bug tracking application, the contents of which you routinely export to Excel to manipulate the data and create unpretentious but thoroughly misleading bar charts so that your managers can sleep better at night and stay off your back during the day. Over time, you become a graph guru and begin to yearn for something more powerful, something that will help you spot anomalies in your data and draw conclusions from otherwise imperciptible patterns. You buy Tableau.

One day, you draw a chart that contains a group of bugs that are related by a pattern you never would have noticed without a tool like Tableau. The bugs all belong to one tenured member of your team. Some of them are resolved and some are active. All of the bugs have been resolved and re-opened more than 10 times, all of them are DCOM-related, all are marked "low priority", and all are stored in different technology buckets. Fishy. You open one of the resolved bugs, inspect it, and realize that it has not actually been fixed. You open another one of the resolved bugs to discover the same thing. Rather than scream across the cubicle wall at your tester for being such a sneaky monkey, you resolve to maintain composure and change the status for all of the resolved bugs in this group to active. Wouldn't it be nice if you could perform that task that in Tableau? Select Region|right-click|Change Value|Field=Status|Value=Active|Update All.

I failed to ask Jock if Tableau is programmatically extensible to the point of enabling a user to hook up a custom Update method. Whether it is or is not, Tableau is pretty sweet. It costs around $1000.

Unfortunately, Tableau doesn't really help me because it is primarily focused on relational data. I am more interested in what I call sub-relational datasets, like wikis, which are a bit more challenging to visualize.

Data visualization tools can dramatically simplify the navigation and administration of WikiWiki.

  • Administration: graphs, charts, and maps can be used as an interface for identifying, monitoring, and moderating acts of WikiSpam, WikiVandalism, and particularly bit-flipping. It can also be used to identify and encourage "healthy" Wiki: practices that maximize the discoverability of information for all users and diminishes duplicate content.
  • Navigation: interactive visualizations can provide a dynamic, realtime, semantically-meaningful alternative to the traditional tree grid and its sexy but under-powered progeny, the hyperbolic tree grid.

Note to self. Must read: