Ananya Devaraju

Every data team knows the sinking feeling when someone asks "why has this number changed?" and the honest answer is "because we fixed something." For Premier Inn's commercial analytics team, that feeling became a recurring reality when a long-overdue platform migration quietly dismantled the data foundations powering pricing, revenue management, and executive reporting across hundreds of hotels. This talk is about what happens when modern architectural patterns meet the demands of commercial analytics — and fall short. Event-driven pipelines, widely recommended as scalable and flexible, turned out to rewrite history with every upstream fix. Metrics shifted, drill-downs contradicted top-line numbers, and the team found themselves measuring data repairs rather than genuine business performance. So they rebuilt. This talk shares how they redesigned the platform around a single principle: reproducibility. A file-based architecture built to preserve historical integrity, tolerate upstream failures, and ensure that a number a commercial team acted on yesterday means the same thing today. If your data platform serves fast-moving decision makers who need to trust what they're looking at, there's something in this talk for you.