Fifty-fifty overall. Until we looked at who was actually buying.
PistonHeads had a core search decision to make: should users search by a single car model or by several? The A/B test came back inconclusive. With no clear signal, the team was stuck on a business-critical feature.
I reframed the question from "which version wins?" to "which users matter most, and what do they need?" I designed a survey that separated what users said they preferred from how they actually behaved, then cross-referenced it with site usage and sales data to profile the highest-value segment. I presented the findings to the product team as a clear recommendation.
The test had averaged everyone together. Split by value, the picture was clear: users who wanted single-model search were more experienced on the site, more likely to buy and worth more over time. Designing for them, rather than for the average user, gave the product team a direction it could defend with commercial evidence rather than opinion.
This is the work of the Fractional Insight Lead: when the numbers alone can't settle a decision, find out whose behaviour matters most.
Tell me where you're starting from. A short note is enough.
