| Authors: | James Socol |
|---|---|
| Time: | 14:00 |
| Session: | http://docs.writethedocs.org/en/2013/conference/talks.html#james-socol |
| Link: | @jasmessocal |
Not a writer, but works on MDN and SUMO at Mozilla. SUMO hosts documentation about Firefox itself, MDN hosts documentation for developers. Historically SUMO was defined by the visual design used for all Mozilla web properties, not a design for a support site. One redesign changed things dramatically, but not clear for the better. There were some symptoms that told the support team that the site was not as useful as it could be. There were low “helpful” scores, high exits from searches, high numbers of search refinement, and high bounce rates.
Around that time someone started working at Mozilla dedicated to usability on the web (as opposed to the browser). SUMO was one of the first things she started working on, because it’s a [potentially] high engagement property. As they analyzed SUMO they found that “if people landed on the right article, the helpfulness score were higher!” :)
First they analyzed the content on SUMO to see what’s linking where, etc. And then they did a card sort of the content, trying to figure out what buckets with what. At the same time they were doing some work to analyze the performance of the current IA, having users do tasks (“where would you find the answer to this?”). The success rates were bad: some as low as 1%. They found that some articles were missing, some were badly named (inconsistent naming, etc), articles that needed consolidation, and “strange attractors”: articles that look right, but aren’t, and you don’t really know it until the end.
Two proposed replacement architectures: one product based, and the other task based. When they tested these replacement approaches using Tree Jack, the results were shocking. Tasks that previously tested at 1% success went to 86%.
After testing, they built wireframes and then updated SUMO. The helpfulness votes went up dramatically after release. “The graph goes the right way: up.”
Interesting that they found the rule about “7 links, 3 levels deep” did not apply to SUMO: if they had the framework right, people could handle a lot more links and information.