Metrics question
Amazon is developing a machine learning algorithm to better target shoppers who have saved items in their shopping carts. The algorithm will send an email to users to remind them about their shopping carts. What factors would you consider when designing this algorithm? What variables would you tweak to yield higher shopping cart purchase rates?
- Amazon
- Metrics
- Hard
Practice this question out loud. An AI interviewer asks it, follows up like a real interviewer would, and scores your answer. Type or speak.
Start a mock interview on this question · Mock interview from a job description
What this question tests
Tests reasoning about a real algorithm design problem: what factors and tunable variables actually drive purchase behavior, balanced against user trust.
How to approach it
- Identify input factors: cart item price and category, time since abandonment, user's historical purchase frequency, and whether the item is in stock or price-changed.
- Identify personalization factors: past response to similar emails (open/click history) and preferred send-time based on user timezone and activity pattern.
- Identify tunable variables: timing (hours after abandonment), frequency (how many reminder emails before stopping), and incentive (whether to include a discount and how much).
- Flag the tradeoff: aggressive discounting raises short-term conversion but trains users to abandon carts expecting a coupon, hurting margin long-term.
- Propose measurement: A/B test timing and incentive combinations, tracking cart-recovery rate and, critically, incremental revenue net of discount cost and cannibalization.
What a strong answer includes
- Separates targeting factors (who to send to, when) from tunable levers (timing, frequency, incentive), which is the structure this question is really asking for.
- Flags the discount-training risk explicitly, a well-known real pitfall in cart-recovery email programs.
- Uses an illustrative number, e.g. assume a 10% discount lifts recovery rate 8 points but erodes margin by more than that on high-frequency buyers, arguing for a smaller or no discount for loyal customers.
- Proposes measuring incremental, not just gross, purchase rate, since some of those purchases would have happened anyway without the email.
Common mistakes
- Proposing to always include a discount without considering the long-term behavior-training risk.
- Ignoring frequency capping, which risks the email becoming spam and hurting the broader email channel's deliverability and trust.
Likely follow-up questions
- How would you measure incremental versus cannibalized purchases from this campaign?
- How would you cap frequency to avoid this becoming a spam signal?
More metrics questions
- Walmart's order return rate is increasing. As a product manager, what things would you look into to isolate the problem?Amazon · Metrics · Medium
- One of the brands in the grocery category is not selling and has stocks left that are soon to expire. Specify all possible reasons for it and also suggest solutions for each of the reasons.Shopify · Metrics · Medium
- A metric for a video streaming service dropped by 80%. What do you do?Google · Metrics · Medium
- You are a PM for an e-commerce company. Due to some mistakes, the data team lost all key metrics. What are the metrics you will track as PM?Amazon · Metrics · Medium
- Your new feature boosts Amazon Search by 10%, adds 2s to load time. What do you do?Amazon · Metrics · Hard
- How would you change the Amazon Fresh experience and how would you measure them?Amazon · Metrics · Medium
More questions from Amazon
Learn the skill behind it
Chapters of the AI PM course that teach what this question tests.
- Chapter 9: Prove it paid off: outcomes, economics, and pricing
- Chapter 2: Data fluency: SQL, logs, and reading the truth yourself
- Chapter 14: Get the job: the AI PM interview loop