Barely two weeks after its launch, Hunch is already taking stock of its progress. “How well do we think Hunch is working?,” asks co-founder Chris Dixon. “Pretty well, although we still have a ways to go.”
Writing on the Hunch blog, Dixon explains that the decision engine is using what it calls “success rate” as its best measure of Hunch’s accuracy. Hunch defines success as when a user clicks “Yes” on one of the top three results and doesn’t click “no” one one of them. Dixon says the site’s success rate is at 81% now, up from 70% when the private beta site was launched. Their goal is to get the success rate to 95% or higher.
Dixon also shares a couple other Hunch statistics:
- Users have answered about 20 million “Teach Hunch About You” questions
- There are more than 63,000 search/decision results in the system
- Users have left 3.8 million feedbacks on those 63,000 results
That last number suggests a pretty solid level of engagement among Hunch users — about 62 feedbacks per result.
Dixon also announced a new feature in Hunch’s system that asks users to prioritize their answers when two or more are in conflict. The example given is when a user says s/he wants an SUV, but doesn’t want to spend more than $18,000. Since you can’t have both, Hunch now asks the user to choose a priority so it can find a better set of suggestions.