Georges

I am Georges Duverger.

Product at Phosphorus (computational genomics). Previously Engineering at eBay and Hunch (acquired). Maker of Fitmeal (1st NLP for food). Taller in real life.

Follow me on Twitter, LinkedIn, Dribbble, GitHub, AngelList, or email.

Ask “Do You Like X?”, Not “What Do You Like?”

I recently sat in a user testing session for a project I am working on at eBay. This new product initially prompted participants to add their interests (e.g., handbags, comic books) to a list. This task turned out to be more difficult than we anticipated.

To collect a list of interests from someone you can:

  1. Infer his/her interests from other sources (like Twitter or Facebook)
  2. Ask him/her directly (“what do you like?”)
  3. Suggest interests and let him/her confirm or deny (“do you like x?”)

Hunch has been very successful using the first option for its Twitter Predictor Game (the average correct score is about 85%). It wasn’t an option on this project, so we started with #2 which we thought was the simpler approach. Suprisingly, participants had difficulties coming up with new interests, or at least troubles expressing them at the right level of granularity. We then experimented with #3 and made an interesting discovery.

It appears to be easier for users to remove interests from a pre-populated list of suggestions than to think of new ones to add. In addition, when participants saw interests they did not like, they intuitively tried to remove them—without being prompted to. I attribute that behavior to a fear of being misrepresented and to an intrinsic sense of order.

Ask “Do You Like X?”, Not “What Do You Like?”

The drawback of that approach is that it makes users more conscious of the accuracy of those suggestions. If they are too far off, this can result in a lesser opinion of the product.

July 19
Blog inspired by Oliver Reichenstein, Dustin Curtis, Chris Dixon, and Ryan Singer. Code forked from David Albert.