Product Intuition: How Senior PMs Make Good Calls With Incomplete Information
Product intuition isn't a gut feeling — it's calibrated judgment. Here's the three-level evidence framework senior PMs use to make confident calls when the data is incomplete.
Share
Product Intuition: How Senior PMs Make Good Calls With Incomplete Information
There's a moment every product manager knows.
The data is inconclusive. The stakeholders want an answer by Friday. Your discovery is two weeks in, not two months. And you have to decide whether to build the thing, kill it, or buy more time — knowing that each choice has a cost.
Junior PMs freeze here. They ask for more research. They escalate. They run a survey that takes three weeks and answers the wrong question.
Senior PMs decide.
Not because they have more data. They often have less. It's because they've developed something that almost nobody in product management talks about directly: product intuition. The ability to read incomplete signals, weight them correctly, and commit to a direction with enough confidence to move — while staying genuinely open to being wrong.
This isn't a soft skill. It's a learnable one. Here's how it actually works.
What Product Intuition Is (and Isn't)
Product intuition is not a gut feeling. Gut feelings are pattern-matching from past experience applied without conscious examination. They're fast and often wrong, especially in new markets or with unfamiliar user segments.
Product intuition is calibrated judgment. It's the output of training yourself to know:
What kind of evidence you're looking at
How much that evidence is worth
What additional signal would actually change your mind
When you have enough to decide and when you genuinely don't
The clearest sign of good product intuition isn't that someone is always right. It's that they're rarely surprised when they're wrong. They knew the uncertainty was there. They made the call anyway because the cost of waiting was higher than the cost of being wrong.
The Three Evidence Levels
Every piece of signal you collect sits at one of three levels. Learning to categorise evidence quickly is the core skill.
Level 1: Proxy evidence
Proxy evidence is indirect. It suggests something without proving it. Examples: a support ticket pattern, a drop in a funnel metric, a sales objection that comes up three times in a month, a user quoting a workaround they've built.
S
Specky Team
Writing about AI-native product development at Specky.
Proxy evidence is cheap to collect and easy to misread. It points at a problem without explaining it. Acting on proxy evidence alone is how you build the wrong solution to the right problem.
Proxy evidence tells you that something is happening. It doesn't tell you why.
Level 2: Observational evidence
Observational evidence comes from watching users interact with your product or close substitutes. Customer interviews (especially switch interviews), session recordings, usability tests, and competitor-feature analysis sit here.
Observational evidence tells you why something is happening. It's much harder to collect than proxy evidence and much more expensive to ignore. When observational evidence contradicts what your proxy signals suggest, the observation almost always wins.
Most PMs underinvest in observational evidence and overinvest in proxy evidence.
Level 3: Causal evidence
Causal evidence comes from controlled experiments. A/B tests, feature flag rollouts with holdbacks, before/after measurements with a clean baseline. This is the only evidence that tells you what will happen if you change something.
Causal evidence is expensive to generate. It requires a big enough sample, a clean test design, and usually several weeks of runtime. Most decisions don't need it.
The mistake most teams make is trying to get causal evidence for everything — running an experiment when a user interview would have answered the question faster and cheaper.
How to Know Which Level You Need
The decision depends on two things: the reversibility of the choice and the magnitude of the investment.
Low reversibility + high investment: you need causal evidence. You're making a bet that's hard to unwind and expensive to get wrong. Run the experiment. Design the holdback. Get the number.
High reversibility + moderate investment: observational evidence is enough. Build a lightweight version, watch users interact with it, decide whether to go deeper. Feature flags were invented for this.
High reversibility + low investment: proxy evidence plus a bias to action. If you're wrong, you'll know quickly and can correct. The cost of waiting exceeds the cost of a small wrong turn.
Most day-to-day product decisions sit in the third category. Most product teams treat them like the first.
The Calibration Habit
Product intuition gets sharper through one specific practice: tracking your predictions.
Before making a decision, write down what you expect to happen. Not a wish — a falsifiable prediction. "If we ship this change, day-30 retention will improve by at least 5% in the cohort that sees it." Or: "Three months after launch, fewer than 10% of users will use this feature more than once."
After enough time has passed to know, come back and check.
This is uncomfortable. Most PMs avoid it because predictions create accountability. But it's the only feedback loop that actually sharpens calibration. You learn which types of decisions you call well and which types you consistently misprice.
Senior PMs who are honest about this tend to find the same two patterns:
They overestimate adoption of features their team is excited about
They underestimate the persistence of current behaviour
Knowing your specific failure modes is worth more than any framework.
Reading Weak Signals Well
Strong signals are easy. A 40% drop in activation is obvious. A pattern that shows up across eight customer interviews is obvious. Product intuition isn't about reading those.
It's about reading weak signals: the single customer who describes an unusual workflow, the slight drop in a secondary metric that doesn't fit any current hypothesis, the sales rep who says "customers don't mention this as a blocker, but they always ask about it anyway."
Three practices improve weak-signal reading:
Separate observation from interpretation. When a customer says "this is confusing," that's an observation. "They want a simpler UI" is an interpretation — and it may be wrong. The confusion might be about the value proposition, not the interface. Force yourself to hold the raw observation before you reach for the explanation.
Look for convergence across sources. A single weak signal means almost nothing. The same pattern showing up in support data, a customer interview, and a sales conversation in the same week is worth investigating. Weak signals compound.
Ask what would make you update. Before deciding what a weak signal means, ask: what evidence would change my interpretation? If nothing would, you're not reasoning — you're rationalising.
When to Decide Without Enough Data
Here's the uncomfortable reality: you will never have enough data. There is always more evidence you could collect. The question is whether the expected value of additional signal exceeds the cost of the delay.
It usually doesn't.
The cost of delay is underpriced in most product organisations. Engineers wait. Designers iterate on something that hasn't been validated. Engineers and PMs discuss the thing instead of shipping and learning. Competitors move.
The right question is not "do I have enough data?" It's "what is the cheapest test that would change my mind?"
If the answer is "nothing would" — you've decided, you're just avoiding accountability for the decision. Make the call and own it.
If the answer is a two-week experiment, run it. If it's a one-hour customer call, make it. If it's a proxy metric you can check in five minutes, check it.
But if the cheapest test that would change your mind is three months of causal data — and the decision is low-stakes and reversible — decide now and use the three months learning from reality instead.
The Senior PM Reflex
The signature of good product intuition isn't confidence. It's something quieter: a clear sense of how much you know, what kind of evidence you'd need to be more sure, and whether it's worth getting that evidence before you move.
It sounds like this: "I'm about 65% confident this is the right call. The thing that would move that to 85% is one more round of customer interviews on the activation flow. I think we should do the interviews. If they confirm what I'm seeing, we build. If they contradict it, we need to understand why before we invest."
That's not a hedge. That's someone who knows exactly where they are in the evidence stack and what the next right move is.
The 59% of PMs who told Pragmatic Institute that strategy is the most important skill for the next three years are right — but strategy without calibrated judgment is just slides. The judgment is what makes the strategy real.
Develop the habit of making predictions and checking them. Train yourself to separate observation from interpretation. Price the cost of delay honestly. And when you're wrong — which you will be, regularly — make sure you were wrong in an informative way, not an avoidable one.
That's the job. The data is always incomplete. The call still has to be made.
Specky is the AI product workspace that shows its work — it connects your research, customer interviews, experiments, and outcomes into a living product graph, so your intuition is built on evidence you can point to, not just instinct you can't explain. → specky.ai