### Prompt On one specific day, the **number** of successful deliveries in a large metropolitan market drops sharply. The alert is about the count, not the success percentage. Diagnose the change and explain how you would decide whether it requires a product or operational response. After your initial diagnosis, suppose a follow-up alert shows that the order conversion rate for one merchant also fell that day. Explain how you would investigate whether the two observations share a cause. ### Constraints & Assumptions - A successful delivery is an order that was placed and ultimately delivered; confirm the production definition before querying. - No cause is given. Customer demand, merchant operations, courier supply, marketplace policies, competition, external events, and instrumentation are all hypotheses to test. - The one-day scope makes seasonality, holidays, local events, weather, outages, and incomplete data especially important. - You may request additional data, but distinguish available evidence from proposed logging. - Avoid claiming causality from a coincident before-and-after comparison. ### Clarifying Questions to Ask - How large is the drop, when was the alert generated, and is the day complete in every source and timezone? - What comparison created the alert: prior day, same weekday, forecast, or control market? - Did total sessions, checkout attempts, placed orders, or delivery success rate change? - Is the decline broad or concentrated by hour, neighborhood, merchant, cuisine, customer cohort, or platform? - Were definitions, pipelines, releases, fees, promotions, merchant availability, or courier policies changed? - For the merchant conversion metric, what is its denominator and where in the funnel is conversion measured? ### Part 1: Decompose the Absolute Drop Build a metric tree that separates changes in demand from changes in funnel conversion and fulfillment. State the first checks and cuts you would run. #### Hints An unchanged delivery success rate can coexist with fewer successful deliveries if fewer orders enter the funnel. Reconcile counts before explaining ratios. #### What This Part Should Cover ```premium-lock What This Part Should Cover ``` ### Part 2: Test Customer, Merchant, Courier, and External Hypotheses Prioritize plausible causes, identify observable evidence for and against each, and explain how you would avoid a long unranked list. #### Hints Use the stage and segment of the break to prune the hypothesis tree. Compare with unaffected cohorts, markets, or time windows where defensible. #### What This Part Should Cover ```premium-lock What This Part Should Cover ``` ### Part 3: Investigate the Merchant Conversion Follow-up Define the merchant funnel, localize the conversion loss, and decide whether the merchant explains a meaningful share of the market-wide delivery decline. Propose the next action or experiment. #### Hints The merchant's rate can fall because its own funnel worsened or because the mix of visitors changed. Quantify both the rate effect and the traffic-volume effect. #### What This Part Should Cover ```premium-lock What This Part Should Cover ``` ### What a Strong Answer Covers ```premium-lock What a Strong Answer Covers ``` ### Follow-up Questions 1. How would you estimate each component's contribution when several funnel stages changed at once? 2. What if placed orders fell but sessions and checkout conversion were unchanged? 3. What if the success rate improved while the number of successful deliveries fell? 4. How would you distinguish a merchant-specific issue from a shift toward lower-intent visitors? 5. When would you run an experiment instead of making an immediate operational fix?