How would you diagnose a completed orders drop?

## Case: Completed orders dropped in Los Angeles You are a Data Scientist supporting a **consumer pricing** team for a two-sided delivery marketplace (customers place orders; merchants and couriers fulfill them). In the last week, stakeholders report that **completed orders in Los Angeles (LA) dropped materially** versus the prior baseline. ### Your task 1. **Clarify and quantify the problem** - Define “completed order” and the exact time window. - Specify the comparison baseline (WoW, YoY, rolling average) and the magnitude of the drop. - Confirm scope: Is it **only LA** or also nearby cities/regions? Only certain order types (e.g., scheduled, grocery) or platforms (iOS/Android/web)? 2. **Define success metrics** - Propose a set of **primary**, **diagnostic**, and **guardrail** metrics relevant to completed orders. 3. **Form hypotheses and a debugging plan** - Lay out a structured set of hypotheses for why completed orders could drop. - For each hypothesis, describe what data you would pull, what cuts/segments you would check, and what patterns would confirm/refute it. 4. **Recommend fixes and experiments** - Propose short-term mitigations and longer-term experiments to recover completed orders. - Describe how you would design experiments (or quasi-experiments) given marketplace and pricing constraints. ### Output Provide a clear, step-by-step approach, including example metric definitions and at least a few concrete experiment ideas.