Amazon Product Manager Interview Questions
Product managers at Amazon work on a store used by shoppers and by outside sellers, on delivery, and on products such as Prime, Kindle and Alexa. Prepare to start every answer from the customer, to explain how you would measure success, to weigh the needs of shoppers against those of sellers, and to tell stories about hard choices you made and what you learned when a launch went badly. What to show here: starting from the customer and owning hard trade-offs. Each question below has what the interviewer is checking, an answer frame and a sample spoken answer to adapt.
How Amazon hires, in its own words, and its questions for every job →
1. How would you improve the experience of returning an item for a shopper who no longer has the original box?
- Describe the customer and the moment of pain.
- Map the steps of a return today and where the box matters.
- Offer a few solutions and choose one.
- Say how you would measure it and guard against abuse.
“I would start with the customer: they have an item to send back, no box, no printer, and they feel it is going to be a chore. I would map the steps and find where the box is truly required. One option is drop-off points that accept the bare item with a code shown on the phone and pack it there. The app should show the nearest points and how long the refund takes once the item is scanned. I would measure how many started returns get completed and how customers rate the process. I would also watch for misuse, such as wrong items returned.”
Weak answers redesign the returns screen in the app and ignore the physical steps the customer must take.
2. Imagine a feature that lets a household reorder everyday items by asking Alexa. What could go wrong, and how would you design for it?
- Describe who uses it and when.
- List what can go wrong: wrong item, wrong size, a child ordering, surprise price.
- Design confirmations and limits that stay light.
- Define success and a safe first launch.
“The customer is someone with full hands who has just run out of washing powder. Voice is good for that moment but poor for showing detail. Things that can go wrong: the wrong product or size, a price that has changed since last time, a child or a television placing an order, or a duplicate order from another family member. I would limit voice reorders to items bought before, read back the item and price, and send a phone notice with an easy cancel window. I would launch with a small set of product types first, and measure cancelled and returned voice orders.”
Weak answers describe only the happy path and do not mention wrong orders or who in the home can order.
3. The store adds the option to collect parcels from a pick-up locker instead of home delivery. How would you measure whether it is a success?
- State the customer problem it solves.
- Choose a main measure tied to that problem.
- Add supporting and guardrail measures.
- Say what result would make you change or stop it.
“The problem being solved is missed deliveries and parcels that are not safe at the door. So my main measure would be successful first-time handover for customers who choose lockers, compared with similar customers on home delivery. Supporting measures would be how many customers choose the option again, and how long parcels wait before collection. Guardrails would be lockers being full when a parcel arrives, and parcels returned because nobody collected them. I would also check whether lockers bring orders from customers who ordered less before. If people try it once and do not return, I would study why before adding more locations.”
Weak answers count the number of lockers installed or total parcels and never ask whether customers came back.
4. Sellers can pay to have their products shown higher in search results. How would you decide how many sponsored results to show on a page?
- Name the three parties: shoppers, sellers, the business.
- Describe the harm of too many or poorly matched sponsored results.
- Propose rules on relevance, labels and number.
- Test with long-term measures, not only clicks.
“Three groups are involved: shoppers who want the best product, sellers who want to be seen, and the business that earns from placements. If sponsored results crowd the page or do not match the search, shoppers stop trusting search, and that harms everyone. I would require a sponsored result to be relevant enough to deserve its place, label it clearly, and cap the number per page. To set the cap, I would test different levels and measure not only clicks but whether shoppers found and kept what they bought, and whether they searched again later. I would choose the long-term result.”
Weak answers maximise the number of paid placements, or refuse them entirely without weighing the sellers' need to be found.
5. Tell me about a time you said no to a feature an important partner or stakeholder wanted because it would have cost customers' trust.
- Describe the request and who made it.
- Explain the risk to customers.
- Show how you said no and what you offered instead.
- Share the result.
“On your project, a large partner asked us to hide reviews below a certain star level on their product pages. I understood their worry, but customers rely on seeing the bad with the good. I gathered what we knew: customers who read negative reviews returned fewer items. I met the partner, shared that, and said we would not hide reviews. I offered something else: a way to reply publicly to reviews and to flag ones that broke the rules. They were not happy at first, but they used the reply feature. Over time, your result. The relationship survived because I came with an option.”
Weak answers give in to the stakeholder, or refuse bluntly without understanding the need behind the request.
6. Describe a launch of yours that did not work. How did you find out, and what did you do about it?
- Describe the launch and what you expected.
- Explain how you learned it was not working.
- Say what you did: fix, change or stop.
- Share what you do differently now.
“I launched a wish list sharing feature on your project, expecting customers to share lists before gift seasons. After a month, few lists were shared and almost nobody bought from a shared list. I spoke with some customers and learned that people felt awkward sending a list of things they wanted. I had tested whether the feature worked, but not whether people wanted it. I proposed we stop investing, keep it running for those who used it, and move the team to another problem. It was uncomfortable to say in a review. Now I test the need with a rough version first.”
Weak answers blame engineering, timing or marketing, or describe a failure that was secretly a success.
7. You see a sudden rise in five-star reviews that look fake for products in one category. What do you do?
- Confirm the pattern with data and examples.
- Limit harm: hold suspect reviews from the ratings.
- Work with the teams that handle abuse and sellers.
- Plan a longer fix and how to measure trust.
“Reviews only help if customers believe them, so I would move quickly but with evidence. I would ask an analyst to confirm the pattern: many reviews in a short time, from new accounts, with similar wording. While that is checked, I would ask for suspect reviews to be held out of the star rating, which can be undone if we are wrong. I would bring in the people who handle abuse and seller policy, since sellers may be behind it or may be victims of a rival. For the longer term, I would look at giving more weight to reviews from confirmed purchases.”
Weak answers delete reviews on a hunch, or wait for a complete analysis while shoppers keep buying on false ratings.
8. Amazon's products serve shoppers, sellers and readers, and reach into homes through devices. Why do you want to be a product manager here?
- Say what kind of product problems you like.
- Link them to the store, sellers or devices.
- Give an experience that shows it.
- Say how you want to grow.
“I am drawn to products where the promise is made on a screen but kept in the physical world. A delivery date is a product decision, and so is what happens when a parcel is late. I also like that many decisions here have two customers at once, the shopper and the seller, and the right answer has to work for both. In my current role I worked on your project, and the part I enjoyed most was following a customer problem through operations. I want to get better at writing down my thinking clearly and at making decisions backed by evidence.”
Weak answers could be given at any large company and never mention customers, sellers or deliveries.
No figures on this page: pay depends on the level, the city and the offer in front of you. When the offer comes, the free salary negotiation script helps with that part.
The questions above are the prep. On the real call, ClapAssist helps with the ones you did not practice.
ClapAssist is an AI interview assistant for Mac and Windows. It listens to the interview on your computer and shows you what to say, in short lines you can read while you talk. Your live interview audio and screen are never stored. Your resume and notes are saved to your account so the app fills them in on any computer. It is hidden from screen share on every plan, including Free; only you can see it.