Diagnose why lunch converts at 2% vs 10% at other meals on a food delivery app

You are the product manager for a food delivery app. A data analyst tells you that usage spikes three times a day, around breakfast, lunch and dinner. At breakfast and at dinner about 10% of that traffic converts into orders, but at lunch only 2% does. How would you work out what is causing the lunch gap, and what would you do about it? ```hint Find where lunch falls behind Before listing causes, break the path from opening the app to a confirmed order into steps and compare each step at lunch with the same step at breakfast and dinner. Then compare who is using the app at each meal and in what situation. ``` ### Clarifying Questions - How is conversion defined: orders per session, per app open, per search or per user active in the window? And is the spike measured in sessions, users or searches? - Has lunch always converted at around 2%, or did it fall recently while breakfast and dinner held near 10%? - Does the gap look the same on every platform and in every city, on weekdays and weekends, and for new and returning customers? - Are lunchtime visitors mostly the same people who order at breakfast and dinner, or a different group? ### What a Strong Answer Covers - Validating the numbers first: the metric's definition and denominator, how the meal windows are drawn, and whether the gap is a tracking artifact - A funnel breakdown that pinpoints where lunch sessions drop off, segmented by platform, location, day type and customer type - Hypotheses on both sides of the marketplace (restaurant and courier supply, delivery times and fees, versus customers' time, setting, intent and alternatives), tested in a deliberate order - A judgment on whether the gap is unmet demand worth recovering or simply a different kind of visit, and which metric fits each case - Fixes tied to the confirmed cause, with a success metric, guardrails and a test design ### Follow-up Questions - You discover that many lunch sessions come from people deciding what to have for dinner, who order later that day. Does the 2% still matter, and how would you change the metric? - Couriers are scarce at noon, so lunch delivery estimates are much longer. Which levers would you pull, and what does each one cost? - How would you test a lunch-only change, such as scheduled delivery to office buildings, without distorting breakfast and dinner results?