A/B Test

Minimizing support costs through content-only experiment

Context: On the Global Content Team at Uber, I was in charge of all help content for riders. A major department goal for the quarter was to address the issues around cancellation fees (high volume, high cost, user frustration).

Problem: Many users who are frustrated by a cancellation fee can’t find quick help (and that costs a lot!).

Solution: Employ a content-only A/B test to prove that changing the way that help options appear to riders would significantly improve the business and customer pain points. For the test group, I created:

  • A tailored help center view with limited options that appeared after a canceled trip

  • A highlight feature for the automated “self-service” that reverses cancellation fees quickly in most cases. Because this feature was hidden, many users had missed it and sought help from an agent for their simple request.

Constraints: No eng resources for the test. Once the idea was proved out, I worked with eng and other partners to improve the self-service and make permanent changes (you can see that result my “Automated Support” project feature).

Results: 10bps drop in support ticket volume on this issue, increased satisfaction among customers