Investigate and Respond to an 80% Collapse in a Streaming Service KPI

You are the product manager for a video streaming service, and you learn that one of the product's metrics has fallen by 80%. Walk the interviewer through what you would do: what you would establish first, how you would find the cause, and how you would respond once you know it. ```hint Size the shock An 80% move is extreme. Think about which kinds of causes could produce a change that large, and in what order it is cheapest to rule them out. ``` ### Clarifying Questions - Which metric fell and how is it defined: a count such as subscribers or hours watched, or a ratio such as a conversion rate? Is a lower value bad for it? - Did it fall overnight or over weeks, measured against what baseline, and how large is the change in absolute terms? - Where was the drop spotted: in one dashboard or report, or confirmed in more than one data source? ### What a Strong Answer Covers - Pinning down which metric fell, its definition and direction, the time window and baseline, and the absolute size - Confirming that the drop is real before searching for a cause - A structured way to locate the loss (which users, surfaces, regions or funnel steps carry it), with internal causes separated from external ones - Hypotheses ordered sensibly for a drop of this size and speed, each paired with the evidence that would confirm or rule it out - A response that separates containment and communication from the longer-term fix, and says how recovery would be confirmed ### Follow-up Questions - Suppose the loss is limited to paid sign-ups on mobile, appears in every market, and nothing changed internally. What would you check next, and how would you respond? - If the drop turns out to be a tracking change, what would you put in place so the next false alarm is caught within hours? - How would you decide whether to roll back a recent release before you have proved it caused the drop?