Metrics question

You are a PM at a hyperlocal grocery delivery startup. You see high returns for orders on rice. What could be the root cause? What are your hypothesis? How would you go about solving this?

Practice this question out loud. An AI interviewer asks it, follows up like a real interviewer would, and scores your answer. Type or speak.

What this question tests

Root cause diagnosis for a specific category level quality problem using structured hypothesis generation.

Confirm the scope, check if the high return rate for rice is unique to certain SKUs, regions, or delivery partners, not a data artifact.
Hypothesize supply side causes, wrong variety or brand substitution, damaged or torn packaging, or expired stock.
Hypothesize demand side causes, customers ordering the wrong size or type by mistake due to unclear product photos or descriptions.
Hypothesize fulfillment causes, picking errors at the warehouse, since rice bags can look similar to staff picking quickly.
Prioritize investigation by likely impact, starting with checking return reason codes if customers provide them at return time.

Starts with the return reason data that likely already exists, rather than guessing blindly, since reason codes would quickly narrow the hypothesis list.
Names a specific plausible cause, warehouse picking errors due to similar looking packaging, and proposes a targeted fix like barcode scanning at pick time.
Proposes a way to validate before rolling out a fix broadly, testing the picking fix in one warehouse and comparing return rate against others.

Jumping straight to a fix without first checking existing return reason data.
Naming only one hypothesis instead of considering supply, demand, and fulfillment causes separately.
No plan to validate the fix actually reduces returns before a full rollout.

What would you look at first if the issue turned out to be concentrated in one warehouse?
How would you design a customer survey to capture the real return reason?