We use Confluence at work for internal documentation. It’s kind of like Wikipedia, but only us developers can see and edit it, and the articles are things like How to set up Server x, or How to Test y.
I love it because we previously used long email chains with the answer buried deep within, mixed with dusty old word documents locked into source control. Confluence feels so much more lightweight, nimble, and searchable.
The Problem
Unfortunately over time it’s become a bit of a mess. We have over a hundred pages but there’s not much rhyme or reason as to how to find them. Confluence does have a search function but, believe it or not, people still browse. Also, if you’re adding an article it’s hard to know where to put it.
The Solution
So I decided to run a card sort exercise with my coworkers. My goal was to understand how they thought about these topics then organize the articles in a way that matched that as close as possible.
I scheduled a lunch and learn, invited the whole development team, and…
(Almost) No One Showed Up

Just one of my coworkers and my manager came. I should know by now that now all developers are as into UX as I am 🙂
Luckily, my manager agreed that this was important enough to schedule in and make mandatory, so a couple weeks later, the workshop was on!
The Actual Workshop
I hadn’t run card sorts recently, so I brushed up on my basics. The two main rules that stuck out for me:
- Run with one person at a time
- Never make people sort more than 30-50 cards at a time
I broke both those rules. Because I only had a one-hour workshop, I had everyone do the sorting in three small teams instead of individually. Also, because we had 100 articles, I did the sort with all 100 of them.

After that workshop, each team ended up with several named groups with several cards each.
Analyzing and Testing the Results
I then used a template from Donna Spencer to analyze the results. I’d definitely recommend this template. It’s free and it did the trick.
Each team came up with remarkably similar groups. Based on those groups, I then came up with proposed categories and tested with a quick tree test using TreeJack from OptimalWorkshop.

Based on the tree test, one of the cards was clearly in the wrong category so I moved it. A couple of the others weren’t decisive so I added cross-links, effectively giving those articles two homes instead of just one.
Reflections
I’m very happy I did this. It was nowhere near perfect but it definitely vastly improved the structure of our documentation, showed everyone which articles were available, and introduced card sorting to the team.
Next Time
If I had more time I probably would have done the sorts one-on-one with my teammates instead of doing everyone at once in teams.
I should have used a smaller representative sample of the articles instead of trying to sort them all.
In other words, I should have followed the basic rules of card sorts!
Oh well! You can’t learn the limits if you don’t go past them every now and then, right?!
[…] done card sorts before to help set up a site’s information architecture, but never to synthesize research. The […]
LikeLike