A/B testing is treated as the gold standard of product decision-making, yet most teams are drawing the wrong conclusions from their experiments. In this talk, Supriya Subudhi draws on a decade of experimentation experience — running multi-country tests at the scale of tens of millions of users across consumer tech and fintech — to expose the most common ways teams misinterpret test results. From peeking problems and novelty effects to metric selection bias and Simpson's paradox, this session will walk through real failure modes and the frameworks that fix them. Attendees will leave with a sharper instinct for when to trust their data — and when to question it.