From Customer-Led Development: How to Build What Your Customers Want When AI Can Build Anything — the operating manual for building what customers want when AI can build anything. See every chapter.
Every product manager you interview will tell you they're customer-obsessed. It's on the resume, it's in the cover letter, and it comes out in the first five minutes of the call. Then you hire them, and they build features nobody uses.
They aren't lying. That's what makes this hard.
Why the resume can't tell you
Most PMs have never worked somewhere that measured customer outcomes. So they think they're customer-focused because they're doing what customer-focused looked like at their last company: a quarterly survey, a persona deck, a few sales calls they sat in on.
They're wrong, and they don't know they're wrong. A resume can't show you the gap, because the candidate can't see it either. They'll describe their work with total sincerity in exactly the language you're screening for.
So you have to stop listening to claims and start testing for practice.
The 47-features interview
Picture a candidate whose resume leads with a big number: shipped 47 features last year. It reads like productivity. It's meant to impress you.
Treat it as a red flag.
Ask one follow-up: "Which of those 47 features had the biggest impact on retention?"
Then watch their face. The customer-led PM answers before you finish the sentence. They name the feature, the customer who asked for it, and what happened to churn afterward. They'll probably also tell you which of the 47 they'd delete.
The other candidate pauses. They talk about velocity, about the roadmap, about how the team hit its commitments. They never get to retention, because nobody ever asked them to. Their old company counted output, and they learned to count it too.
One question and you have your answer. "Shipped 47 features" is less often a sign of a great PM than a sign of a feature factory. The number tells you how much they built. It says nothing about whether any of it mattered.
Rule of thumb: Forget how many features a PM shipped. Ask how many are still used, and what each one did for retention.
Three questions that separate claim from practice
The 47-features moment generalizes. Customer-led skills are demonstrable, so you can test them directly. Three questions do most of the work in a first call:
| Ask | The customer-led answer | The warning sign |
|---|---|---|
| "Name three specific customers who influenced your last major decision." | Named companies, named people, what they actually said | "Our enterprise segment," "our core users" |
| "Tell me about a time you were wrong about what customers wanted." | They light up, and have a dozen examples and the lesson from each | Defensiveness, or no example at all |
| "How did your last feature affect retention or expansion?" | A clear line from the work to the business | Talk of launches, adoption, and story points |
The second question is the sharpest. A PM who can't name a time they were wrong about customers has either (a) not talked to enough customers to ever be wrong, or (b) can't admit mistakes. Both are disqualifying.
The tells you'll start noticing
Once you're listening for practice instead of vocabulary, the signals get loud:
- They say "users," never "customers." Users are abstract. Customers imply a relationship.
- All successes, no failures. Either they're shading the truth or they haven't taken enough risks.
- Vanity metrics. DAU, features shipped, and story points measure activity. None of them is an outcome.
- "The data told us." Data matters, but if it's the only thing they mention, they're data-driven, not customer-driven.
- Nothing they chose not to build. Anyone can build things. The best PMs kill ideas before they cost anything.
Watch out for the impressive traps too. Big-company pedigree teaches big-company process, and at a 50-person startup those best practices become your bottlenecks. A big vision with no customer evidence behind it is guessing with confidence. Ask what would prove the vision wrong. If they can't answer, their vision is religion.
The trap of potential
The most expensive mistake comes at the end of the process, when a candidate fails these tests and you hire them anyway. They're sharp. They interview well. They could figure it out.
Potential lets you ignore the evidence in front of you. Either a candidate can name real customers and tell real validation stories, or they can't. Talking to customers isn't a skill people pick up once someone hands them the job. It's a habit they bring with them, or one they don't have.
If they don't already operate customer-led, they won't start just because you hired them.
That's the whole test. Look past what they claim and listen for what they already do. The candidates who pass usually do it without noticing, because to them it's simply how the job works.
The full interview bank, the four-stage hiring process, and where to find PMs who already work this way are in Customer-Led Development.
If you want more tests that cut past what people claim to what they actually do, subscribe to the Context Limit newsletter.