Earning the Opinion

Everything you decided in the last three docs, the persona who matters, the pains that count, the bet that wins, was reasoned from a chair. That makes it a hypothesis, not a fact. Product sense is having an opinion. Being world-class is knowing the difference between an opinion you’ve earned and one you’ve merely argued yourself into, and doing the cheap work to close that gap before you bet the team on it.

This is the step most people skip. It’s more satisfying to have a confident view than to go test it, and confident views are exactly how teams ship the wrong thing with great execution. The move that separates a good product thinker from a dangerous one is the same every time: reduce the uncertainty before you commit the capacity.

Your Opinion Is a Stack of Assumptions#

What’s expected: You can list the assumptions your plan depends on, and say which ones would sink it if they’re wrong.

Any product opinion is really a chain of guesses wearing a trench coat. “We should build one-tap checkout” quietly assumes: that re-typing details is genuinely why shoppers abandon, that it costs us enough orders to matter, that one-tap would actually recover them, that shoppers trust saved payment, that it’s the highest-value thing we could build with that time. Each link can be true or false, and the whole opinion is only as strong as its weakest one.

How to get there: Before you defend a plan, write its assumptions as a list. Then mark each one:

Known — we have real evidence this is true.
Assumed — we believe it, but it’s a guess.
Load-bearing — if this one is false, the whole plan collapses.

The assumptions that are both assumed and load-bearing are your job. That’s the uncertainty worth spending time to reduce, and everything else can wait. You are not trying to prove the whole plan. You’re trying to find the one or two beliefs that, if wrong, waste a quarter, and check those first.

Silence Is a Hypothesis, Not a Fact#

What’s expected: When you infer a problem from the absence of a signal, you treat it as something to measure, not something you already know.

A recurring move in this group is reading meaning into what people don’t do: a shopper who never completes checkout, a visitor who never returns, a feature nobody uses. That instinct is good, silence is often where the biggest problems hide, because the people it hurts don’t complain, they just leave. But silence is ambiguous. People might be abandoning checkout because it’s broken, or because they were only browsing, or because the price scared them. The absence tells you where to look, not what’s wrong.

So an inferred problem is a hypothesis with a confidence level, not a confirmed High-impact item. “Shoppers are abandoning checkout” earns a High ranking only after you’ve looked. Which brings us to the cheap ways to look.

The Ladder of Evidence: Cheap to Expensive#

What’s expected: You reach for the cheapest form of evidence that would actually change your decision, before the expensive ones.

You don’t need a research department. You need to climb only as far up this ladder as it takes to become confident enough for the size of the bet. Most decisions are settled on the first two rungs.

Talk to them

A 20-minute conversation with five real users of the persona you're building for. Cheapest, fastest, and the single highest-yield thing most teams never do. Ask what they were trying to do and where it went wrong, not whether they'd like your idea.

Watch them

Observe someone actually using the product, or watch session recordings. People misreport what they do; watching never lies. You'll see the real friction in minutes.

Measure them

Pull the data. Where exactly do people drop off, how many, how often? To measure a step you first have to instrument it, add tracking so the product records that the step happened. If you can't see it, your first cheap move is to make it visible.

Test it

Ship a small, cheap version to a slice of users and watch what changes, before committing to the full build. A fake door, a prototype, a limited rollout. The most reliable evidence, and still far cheaper than building the whole thing.

The order matters. A conversation costs an afternoon and often kills a bad idea before you’ve spent anything. Instrumentation costs a day and turns “shoppers are abandoning checkout” from a guess into a number, and shows you exactly which step they quit on. A full build costs a quarter. Spend the afternoon before the quarter.

Match the Evidence to the Stakes#

What’s expected: You gather more evidence for bigger, harder-to-reverse bets, and you don’t over-research the small reversible ones.

Earning the opinion is not “research everything.” That’s its own failure, a team that studies a one-day fix for a week has wasted more than the fix costs. The rigor scales with the bet, exactly as in Constraints & Cost:

Small and reversible (a minor tweak): don’t research it. Ship it and watch. The shipping is the test, and it’s cheaper than the meeting about it.
Large and reversible (a real feature you can pull back): a couple of rungs, talk to users, check the data, before you build.
Large and irreversible (a rock, a core rewrite): earn it hard. Conversations and data and a cheap test. This is where “be damn sure” is literal.

The skill is calibration: knowing how much certainty a decision actually needs, then buying exactly that much and no more.

Pressure-Test Your Own Bet#

What’s expected: Before you present a big bet, you’ve tried hard to kill it and can say what evidence would.

The winning bet is the most dangerous thing you’ll propose, because it’s usually the biggest and the one you’re most attached to. So you owe it the harshest treatment. Take your bet and genuinely attack it, the way a skeptical colleague would:

What would have to be true for this to be a mistake? Name the conditions. Then go check whether any of them hold.
What’s the strongest case against it? For “one-tap checkout”: maybe shoppers abandon over the shipping price, not the typing, in which case one-tap changes nothing. Maybe they actively want a review step before paying. If you can’t rule those out, you don’t yet know the bet is real. (This is not hypothetical: the same Storefront’s “one-tap add-on” idea got killed in testing, shoppers read it as pushy and refunds ticked up.)
What single piece of evidence would change my mind? If you can’t name one, you’re not holding an opinion, you’re holding a belief, and beliefs don’t belong on a roadmap. Go get that piece of evidence.

A bet that survives a genuine attempt to kill it is worth committing to. A bet nobody tried to kill is just the loudest idea in the room, and the loudest idea being wrong is how quarters get wasted.

The Standard#

The armchair gets you a hypothesis. That’s a real contribution, most people can’t even do that. But the job doesn’t end there. A world-class product person is defined by what they do next: they find the cheapest evidence that would change the decision, they get it before they spend the team’s capacity, and they can always answer “how do you know?” with something better than “it’s obvious.”

Have the opinion. Then go earn it.

Next: Goals & Success Metrics