The Worst Product You Ever Shipped
One interview question that separates PMs with real reps from PMs with a clean résumé.
- Difficulty
- Easy
- Time to result
- ~days to results
- Steps
- 5
- Confidence
- 90%
Maggie Crowley (VP Product at Toast) asks the same question in every product interview: what's the worst product you ever shipped? Her view is that you're not a good PM if you haven't shipped something genuinely bad — you haven't had enough reps. The signal isn't just that a candidate has a bad ship in their history; it's whether they know which one it is and can admit it. The framework turns that into a scoring rubric, and pairs it with the failure mode she's still angry about: the rewrite.
Origin
Maggie Crowley's standard product interview question, plus her own worst project — an unnamed rewrite estimated at six months that ran two and a half years and never reached feature parity.
Core principles
- 01If a PM has never shipped something really bad, they haven't shipped enough.
- 02The signal is not the failure — it's the ability to name it and own it without prompting.
- 03A candidate who can't identify their worst ship either lacks reps or lacks honesty. Both disqualify.
- 04Rewrites are the canonical PM disaster: red flag number one.
- 05Arrogance plus skipped discovery is the standard recipe.
- 06Once a redesign/rewrite is far enough along, everyone builds in the new world — and you're committed even when the experiment reads negative.
How to run it
- 1
Ask the question in every product interview
Ask: what's the worst product you ever shipped? Ask it of every candidate, every time, so you build a calibrated distribution of answers.
Pro tip Ask it early enough that the candidate hasn't yet learned which parts of their history you value.
- 2
Score the presence of an answer
First gate: do they have one at all? A PM with no bad ship in their history has not had enough time or enough reps. A blank here is a signal about experience, not about excellence.
Pro tip Watch for the deflection into an interview-safe near-miss ('we shipped a bit late'). That's a non-answer.
Watch out Junior candidates may legitimately have no bad ship. Calibrate the bar to the level you're hiring for.
- 3
Score the specificity and the ownership
Second gate: do they know exactly which one it was, and will they admit it? Crowley's own answer names the project, the wrong estimate, the arrogance, the skipped discovery, the missing one-pager. That level of specificity is the signal.
Pro tip Push on the causal chain: what specifically did you skip, and what would you do differently? A real answer has named process failures, not vibes.
- 4
Score the diagnosis quality
Third gate: is their post-mortem structural or self-flattering? Crowley's has a clean structure — (a) we got arrogant and thought we could do it, (b) we skipped discovery: no one-pager, no technical or design research into what the requirements actually were, we just went for it.
Pro tip The strongest answers include the organizational dynamics — sunk cost, people rotating in and out, everyone privately thinking it was dumb.
Watch out Beware the answer that lands on 'and it turned out great in the end'. That's the 'my weakness is perfectionism' move.
- 5
Apply the rewrite rule
As a standing decision rule: don't rewrite. If anyone tells you to do a side-by-side rewrite, don't. The pattern is that the estimate is off by 5x, feature parity never arrives, people rotate through, sunk cost takes over, and by the time the experiment reads negative everyone is already building in the new world — so you launch anyway and spend months clawing back to neutral.
Pro tip The temptation returns every three or four years: 'we haven't updated this thing in a long time, maybe this time the rewrite works'. It doesn't.
Watch out The engineer you trust most, who you've shipped with many times, will confidently say six months. That confidence is not evidence.
In the wild
Crowley's team decided to rewrite a core, long-standing part of the product — the kind where the original code was still the founders'. A trusted engineer with whom she had a strong working relationship estimated six months. It took two and a half years and still wasn't done. People rotated in and out, everyone thought it was dumb, and sunk cost kept it alive. The causes: they got arrogant and thought they could do it, and they skipped discovery — no one-pager, no technical or design research into the actual requirements.
→ It never became a success and did not change the trajectory of the business. Her rule since: don't rewrite. If anyone ever tells you to do a side-by-side rewrite, don't.
Common mistakes
Hiring the PM with no scars
A candidate whose history contains no bad ship hasn't had enough reps. Worse is the one who has had them but can't name them — that's an honesty problem, and it will show up in every post-mortem they ever run for you.
Trusting the confident rewrite estimate
A trusted engineer said six months; it took two and a half years and never reached parity. Estimates for rewriting old, founder-era code are systematically wrong by multiples, because nobody knows what the requirements actually are.
Skipping discovery because you've done it before
No one-pager, no technical or design research into the real requirements, just go. Arrogance about a familiar surface area is precisely what makes the rewrite unbounded.
Passing the point of no return on a redesign
Once everyone is building in the new world, a negative experiment result no longer stops you. You launch anyway and then spend the following months trying to claw back to neutral.
Is it for you?
Best for
Hiring managers building a product team who want a fast read on a candidate's reps, honesty, and post-mortem quality — and PMs facing pressure to greenlight a rewrite.
Not ideal for
Interviewing genuinely entry-level candidates with no shipping history, where the absence of a bad ship carries no signal.
From the transcript
“a question I ask in every product interview is uh what's the worst product you ever shipped”
“I don't think you're a good PM if you haven't shipped something that's really shitty”
“it didn't take six months it took two and a half years it still wasn't done”
“arrogant and we thought we could do it B we skipped Discovery we didn't really write a one pager we just went for it we…”
“don't rewrite if anyone ever tells you to do a rewrite don't do it”
From the episode
Failure