Estimation question

How would you estimate purchase order abuse in Amazon? How would you solve it?

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

Estimation combined with a solution: structuring a fraud-sizing problem with clear assumptions, then proposing a mitigation.

Define purchase order abuse clearly: businesses using Amazon Business purchase orders to obtain goods without paying, or misusing terms for personal gain.
Break the estimate into a funnel: total Amazon Business purchase order volume, times an assumed fraud rate, times average order value.
State assumptions explicitly, for example Amazon Business processes a large volume of B2B orders monthly, and assume a fraud rate around 0.1 to 0.5 percent based on typical B2B credit fraud rates.
Calculate a rough dollar figure from these assumptions and sanity check it against typical retail fraud loss rates of under 1 percent of revenue.
Propose solutions: stricter credit vetting for new business accounts, anomaly detection on order patterns (sudden large orders from new accounts), and delayed shipment holds pending payment verification for flagged accounts.
Prioritize anomaly detection as the highest leverage fix since it catches abuse without adding friction for legitimate repeat business buyers.

States every assumption explicitly and out loud, for example order volume and fraud rate, so the interviewer can challenge specific numbers.
Sanity checks the final estimate against known industry fraud loss benchmarks instead of leaving a number unchallenged.
Pairs the estimate with a concrete, targeted mitigation, anomaly detection on new account order patterns, rather than a generic 'add more verification' answer.

Giving a number with no stated assumptions or logic behind it.
Proposing fraud mitigation so strict it would hurt legitimate business customers.

Chapters of the AI PM course that teach what this question tests.