Description
The first clue was a complaint that sounded trivial: customers said the checkout button had stopped working. The engineering dashboard showed no widespread outage, the payment provider reported normal operation, and total daily sales were only slightly below forecast. Yet customers on a particular mobile browser were leaving the store midway through payment. A descriptive report could have counted the abandoned checkouts. A diagnostic investigation had to explain why the abandonment occurred, identify whose experience had changed, and distinguish a plausible story from a defensible finding. That distinction animates this book. Descriptive analytics establishes what happened; diagnostic analytics investigates the mechanisms, circumstances, and contributing factors behind an observed result. Its everyday questions are deceptively demanding: Why did complaints rise? Which part of a process failed? Did the new software release cause the slowdown? Was a sales decline due to customer mix, pricing, inventory, or the measurement system itself? Good answers depend on evidence that is often incomplete and occasionally misleading. Diagnostic analytics is not simply the art of attaching explanations to charts. It combines data provenance, comparative analysis, process knowledge, statistical modelling, causal thinking, and practical tests. A careful analyst knows that the strongest correlation is not necessarily the most important cause, that a model can explain variation while misrepresenting a mechanism, and that a useful diagnosis may be provisional rather than final. The goal is not certainty at all costs. It is a justified account of what most likely produced the observed outcome and what further evidence could challenge that account.