In the actual user test, the three options above were randomized, so the final option woudln't succeed simply because people get better at the concept
of segmenting users.
There are a lot of rough edges inside the prototypes, but the path to completion of persona segments was a complete one. Any of the persona segments could have been completed with any of the prototypes.
- First Link is a recreation of segment editor we were currently using, which was designed about 2 years prior. This one performed OK as long as it was not first in the test: people could understand it if they had experience building segments, but it was intemidating if it was their first exposure to the concept.
- Second Link is the "buried details" version. Users could potentially make a decent segment without ever dipping down into details, and they could easily read the results at a later time. This was the overall winner, and it served as the foundation of what we implemented.
- Third Link is the "wizard" prototype, where people could modify most of the details of a question with the mouse in a flyout helper. This did astonishingly poorly.