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.