So I sorta slacked on my CHI blogging last week, but there I wanted to make sure that I blogged about a cool talk from a GA Tech student / Google intern regarding the help module in Google’s AdWords.

For those not familiar with Card Sorting, it’s a usability method for determining a site’s architecture. For example, the National Cancer Institute has webpages on a number of cancer topics, each with subpages about causes, treatments, etc. Each subpage is represented on a “card,” which users sort into the hierarchy that makes sense for them.

In this talk, Google mentioned two card sorting phases. One consisted of one large card sort, which was done with one AdWords expert and 3 usability experts collaboratively. The second consisted of three separate card sorts, each done by an AdWords customer service representative with the instructions to behave like a typical AdWords users.

But no actual AdWords users.

I’ve been doing Card Sorts for the Online Information Design class for the past two years, and I’m intrigued by the fact that Google didn’t use actual users. One of the challenges in Card Sorting is making sure that users aren’t reverting to “naive” sorting. In the National Cancer Institute example, a user sorting “naively” might sort all medications into one large group, without thinking that each one might correspond to a particular cancer. Offsetting this behavior is tricky at best - you have to encourage users to ask questions about unclear terms, provide defintions that are simultaneously informative and not revealing of the underlying architecture, etc. In short - it’s tricky.

So, a question - do you really need to rely on actual users for card sorting? Google didn’t, and instead conducted a usability test to compare the old vs. new architectures, finding the new one to be more appropriate. At CMU, we kept getting inundated with “The user is not like me,” but maybe that doesn’t have to be the hard and fast rule at every turn.