Product Manager Interview (+ AI PM) · Estimation & Market Sizing · Lesson 2 of 2
Sanity checks & handling uncertainty
A great estimate ends with a sanity check: does the answer pass a smell test against something you already know? This is where you show judgment, not just arithmetic.
- Cross-check with a known anchor: 'That implies every American buys 3 a day — plausible for coffee, not for cars.'
- Check the order of magnitude: being within ~2–3x is a pass; being 1000x off signals a broken assumption.
- Flag the biggest sensitivity: 'this answer is most sensitive to the replacement-rate assumption' shows you know where the risk is.
- If an assumption is shaky, give a range rather than false precision.
Catch your own errorIf the sanity check fails, say so and fix it live: 'That number seems too high — let me revisit the frequency assumption.' Self-correcting in real time scores higher than defending a wrong number.
◆ Lock it in
- End every estimate with a sanity check against a known anchor.
- Aim for the right order of magnitude; flag the assumption the answer is most sensitive to.
- Self-correct live if the check fails — it signals judgment.
Feynman drill — say it out loudDescribe two ways to sanity-check a market-size estimate you just produced.
Step 1 rate your confidence · Step 2 pick your answer
Your estimate implies the average American buys 50 new cars per year. What's the right move?
Step 1 — how sure are you?