Product manager interviews probe product thinking, comfort with metrics, and how you decide what not to build. Below are the most common questions with model answers. Senior: architecture, trade-offs, mentoring, and decision-making.
From the value it replaces rather than from cost — what the customer spends today on the problem, in money or in hours, sets the ceiling. Then the model matters more than the number: per seat, per usage, or flat changes who can buy and how the account grows. Pricing is also the fastest lever available, and the one teams revisit least.
Through a chain each level can act on: the strategic bet, the outcome that would prove it, the metric that measures the outcome, and the work a team can own this quarter. If a team cannot name which bet their work serves, the decomposition failed, and they will optimise something local. The check is whether a team can decline a request by pointing at the chain.
Establish which part flattened — acquisition, activation, retention or revenue — because each has different causes and remedies. A plateau in acquisition with healthy retention means a market or channel limit; flat retention with strong acquisition means the product is not holding people. The temptation is a redesign, which changes everything and explains nothing.
By who the customer is and what they are contractually required to do. Regulated industries and large enterprises often cannot use shared infrastructure, and that decides it regardless of engineering preference. The cost is real: on-premise means versions in the field, upgrades you do not control, and support of environments you cannot see, so it should be priced accordingly.
Make the trade-off explicit rather than seeking consensus: what each goal is worth, what it costs, and what is given up by choosing the other. Consensus on a genuine trade-off usually means nobody understood it. If it stays deadlocked the decision belongs one level up, and taking it there is not a failure — pretending both are the priority is.
Start from one transaction and work outward: revenue per order, direct cost, the share of orders that need intervention, and the acquisition cost of the customer who placed it. Then the assumptions that dominate — usually frequency and retention — get named and tested first, because the model is only as good as those two. A model whose sensitivity is not known is a spreadsheet, not an argument.
You group users by when they arrived and follow each group over time, instead of averaging everyone together. That separates 'the product is getting better' from 'we are acquiring more people' — an overall retention number can stay flat while every new cohort is worse, because older, healthier cohorts are propping it up. It is also how you see whether a change actually improved anything, since only cohorts after the change should move.
The channels differ less than the maths does: you compare acquisition cost against the value the channel's users actually deliver, per channel and per cohort, because a cheap channel bringing people who never convert is expensive. Judge on payback period as well as ratio — a channel that pays back in three months can be funded from revenue, one that takes eighteen has to be funded from somewhere else. And be honest about attribution: the last click gets credit it did not earn.
In a real interview you speak under pressure. Cam asks these same questions, scores every answer, and shows exactly what to fix.
Practice a Senior Product Manager interview →