Product sense is the competency most tech companies test hardest for and teach least well. It sits alongside execution, analytics, and strategy as one of the four things a serious PM interview probes, and hiring panels weight it heavily because it predicts whether someone will make good calls when the data runs out. Yet almost no one can tell you how they built theirs. They point at years of experience and shrug.
Here is the uncomfortable part. You can ship products for ten years and end that decade with roughly the same product judgment you started with. Time on the job is not the input. What you do with the feedback is.
What product sense actually is
The tidiest working definition comes from Jules Walter, who spent years at Slack and framed product sense as “the skill of consistently being able to craft products (or make changes to existing products) that have the intended impact on their users.” Notice the word “consistently.” Anyone can get one product right by luck. Product sense is the ability to be right more often than chance, before the results are in.
Nielsen Norman Group puts a sharper mechanical frame on it. They describe product sense as the ability to recognize when a current problem matches a past success or failure, and to reliably estimate how a similar solution will move the outcome you care about. In plain terms, it is pattern matching. You have seen enough problems and enough consequences that a new situation lights up a memory of what happened last time, and that memory is usually accurate.
That is not mysticism. It is the same thing an experienced nurse does when she looks at a patient and says something is wrong before the chart confirms it. And there is a large body of research on exactly when that kind of intuition can be trusted and when it is quietly making things up.
The two conditions intuition needs
In 2009, Daniel Kahneman and Gary Klein published a paper in American Psychologist with an unusual origin. Kahneman had spent his career documenting how expert intuition fails. Klein had spent his documenting how it succeeds. They sat down to resolve the disagreement and largely did. Their answer, Conditions for Intuitive Expertise: A Failure to Disagree, lays out two conditions that have to hold before anyone’s gut is worth trusting.
First, the environment has to be high validity. The cues you see have to be genuinely, stably connected to the outcomes you are predicting. A chess board is high validity; the pieces really do tell you who is winning. A stock ticker on a given day is not; the wiggles do not reliably predict tomorrow.
Second, and this is the one PMs skip, you need prolonged practice with rapid and unequivocal feedback. You have to make the call, then find out whether you were right, quickly and clearly enough that the lesson lands. When both conditions hold, intuition is real recognition. When either fails, Kahneman and Klein are blunt about what you get instead: substitution and overconfidence. You feel certain, and your certainty is manufactured.
Product work is high validity enough. Real user behavior really does connect to design choices in learnable ways. So the first condition is mostly met. The problem is the second one, and most PM careers are structured to violate it.
Most PM careers break both conditions
Think about how the feedback loop actually runs on a product team. You form a hypothesis, you build for a quarter, you ship, and then the interesting part happens slowly. Retention curves need weeks to separate. A pricing change needs a full billing cycle before the churn shows up. The second-order effects, the ones that teach you the most, can take two or three quarters to surface.
Now think about how long the person who made the call stays pointed at that specific bet. Reorgs move PMs to new surfaces. Context switching pulls attention across four initiatives at once. The say-do gap means the early signal often lies to you anyway, so the only honest read is the late one, which arrives after you have already moved on. And the biggest structural break of all is career mobility itself. The median PM tenure at a single company has hovered under two years for most of the last decade.
Do the math on that. If it takes six to nine months to see whether a real bet worked, and you change jobs every eighteen to twenty-four months, you get to close the loop on a small handful of your own decisions before you leave. You shipped constantly. You watched almost none of it land. You accumulated output, not feedback.
I watched a version of this play out long before I worked in software. In IT operations, I signed off on a major network rollout in 2011 whose real consequences, the maintenance load, the failure modes, the thing we got subtly wrong, did not become visible for the better part of a year. The engineers who had rotated off the project by then never learned what their choices cost. The ones who stayed got a permanent upgrade to their judgment. Same work, same intelligence. The only difference was who was still in the room when the results came in.
Closing the loop on purpose
If product sense is pattern recognition and patterns only form when you see the outcome, then building it is not about doing more. It is about deliberately closing loops that your job would otherwise leave open. A few practices actually move the needle.
Write your prediction down before you ship. Not the goal, the prediction. “I think this will lift activation by four to six points, and I think power users will barely notice.” Specific, quantified, and falsifiable. NN/G makes this the pivot of the whole method: state the expected outcome in advance, then measure the real one, because a hypothesis you never recorded is one you will unconsciously rewrite to match whatever happened. Your memory is a flattering editor. A written prediction is not.
Stay attached to your bets long enough to be graded. When a reorg tries to move you off a launch before the numbers mature, negotiate to keep visibility into it even after you own something new. The three months after everyone stops paying attention is where the actual lesson lives.
Spend real hours watching users, not dashboards. Walter recommends sitting in on research sessions two to four times a month and pulling apart an unfamiliar product for an hour or two on top of that, asking the plain questions Julie Zhuo uses: what is this product trying to be, what does it want me to do, does it deliver on what it promised in the first minute. Dashboards tell you what happened. Watching a real person hesitate tells you why, and why is the part that transfers to the next problem.
Grade your calls out loud with someone senior. Walter’s third practice is learning from strong product thinkers by asking what prompted a decision and what alternatives they weighed. Do the same in reverse on your own bets. A quarterly conversation where you walk through what you predicted, what happened, and where the gap was is worth more than another framework. Frameworks are borrowed patterns. This builds your own.
The tradeoff nobody names
Here is the part that connects straight to your career. The single fastest way to raise your compensation over the past several years has been to change companies every couple of years, and job-hopping’s pay premium has been real money. But the same move that maximizes salary is the one that starves product sense, because you keep leaving before your bets are graded. You are optimizing for the number on the offer and quietly disinvesting in the skill that would eventually get you a bigger one.
That does not mean stop moving. It means know which account you are drawing down. If you leave every eighteen months, you have to work far harder to close loops inside that window, and you have to be honest that the years are not automatically buying you judgment. The senior PM plateau that so many people hit is often this: a decade of shipping, very little of it graded, and a level of product sense that stopped growing around year three because that is when they stopped staying long enough to learn.
The PMs who break past that plateau are rarely the ones with the most launches on the resume. They are the ones who can look at a problem they have never seen and be quietly, unfashionably right about it. That is not a gift. It is a stack of closed loops, most of which they had to fight their own job to close.