Your A/B test shows that removing the onboarding tutorial increases Day-1 activation by 5%. The data says ship it. But you also know from user interviews that the tutorial is the main way enterprise users learn critical features — and enterprise accounts represent 80% of your revenue. The data is right about one metric and blind to the context that matters most.
The phrase "data-driven" has become a badge of honour in product teams. It sounds rigorous. It sounds objective. But taken literally, it means letting data make your decisions — which ignores context, outliers, qualitative insights, and the judgment that experienced PMs bring to the table. The alternative, data-informed, treats data as one essential input alongside expertise and context. The distinction sounds subtle. In practice, it creates fundamentally different cultures.
The Core Idea
Data-driven decision-making means the data determines the outcome. If the numbers say X, you do X. There is no room for overriding the data with judgment. This works well in situations with clear causation, large sample sizes, and well-defined metrics — like optimising ad bidding or ranking search results. Algorithms can make better decisions than humans in these contexts because the variables are known and the feedback loops are fast.