Root-Cause a One-Day 20% Drop in Daily Product Ratings on an E-commerce Marketplace

You are the product manager who owns product ratings on a large e-commerce marketplace such as Flipkart or Amazon. On a typical day buyers leave about 10,000 ratings, but on one day the total came in at 8,000. Work out what caused the drop. ```hint Decompose the count A daily total is the last step of a chain that starts well before anyone gives a rating. Break the count into the steps that produce it, and do not assume the number itself is correct. ``` ### Clarifying Questions - Does "rating" mean star ratings only, or written reviews too, and is a rating counted when it is submitted or only after moderation approves it? - How much does the daily count normally vary, for example between weekdays and weekends or around holidays? - Is the day's data complete, and which time zone defines a day? - Did the drop show up across every platform, region and category, or only in some of them? ### What a Strong Answer Covers - Checking that the 20 percent drop is real: pipeline completeness, logging, the metric's definition, and normal day-to-day variation. - A decomposition of daily ratings into the funnel that produces them, including the lag between delivery and rating. - Internal causes (releases, experiments, prompts, filters) separated from external ones (holidays, sale events, outages), checked in a sensible order. - Segment cuts that isolate the cause, and how each hypothesis would be confirmed or ruled out. - Next steps once the cause is found, including whether the drop actually matters. ### Follow-up Questions - The drop appears only on one app platform. What do you check next? - Ratings recover fully the next day. Would you still investigate, and how far would you go? - Volume is flat, but the share of five-star ratings jumped overnight. What could explain it, and why might it matter?