Product sense, without the generic answer

CIRCLES — comprehend, identify the user, report their needs, cut through prioritisation, list solutions, evaluate trade-offs, summarise — is a reasonable checklist. It is also the most over-recited framework in the discipline, to the point that interviewers can hear it coming and some actively discount it.

The problem is not the checklist. It is that candidates deliver the checklist as the answer, announcing each letter like a train station. The framework's job is to stop you forgetting a step, which is a job it does silently.

WatchLewis C. Lin · 10:11

Ten minutes from the person who invented the framework, which is worth more than any secondhand summary of it. Watch it for the scaffolding, then come back — the two steps below are the parts he necessarily leaves as "use your judgement", and they are where the round is actually decided.

Product Design Questions — the CIRCLES Method, explained by its creator Play “Product Design Questions — the CIRCLES Method, explained by its creator” on YouTube

So take the steps as given, and spend this lesson on the two that decide the round.

Step 1: pick one segment, and say why in numbers

Almost every candidate lists segments. Very few choose, and fewer still justify the choice. The listing is worth almost nothing — it's a step everyone completes — while the justification is where taste becomes legible to someone who cannot read your mind.

Here is the honest difficulty: the reason to pick a segment is usually a mix of size, headroom, and strategic fit, and those three point in different directions. So compute them rather than gesturing at them.

"I'll focus on new parents" is a taste claim. Size it and it becomes a decision. Reach × frequency × the gap you can close is the opportunity, in units the business already tracks. Take a product at 40M monthly actives, and three candidate segments — where gap is the additional sessions per month you believe solving their problem unlocks:

Extra sessions per month

Users in the segment

Monthly actives

Share of MAU

Extra sessions per user

the question being decomposed

Every row of that table is this one multiplication. Users in the segment is 40M actives × the segment's share; the opportunity is that count × the gap you believe you can close. Both leaves at the bottom are facts you can look up. The right-hand branch is not — the gap is your estimate, which is why it is the number the interviewer will push on, and why "+5 sessions for new parents" needs a sentence of justification that "4% of MAU" does not.

Read that as an argument you could have in a room, because it is one.

On raw lift, commuters win — 17.6M against 8.0M, 2.2× the additional sessions of the new-parents segment — and occasional users are bigger still. If the only thing you can say is "new parents feel underserved," you lose this argument to anyone holding a calculator.

The counter-argument is in the per-user column, which is the same numbers divided differently:

A five-session gap on a three-session base is a 167% change in how those people use the product. The same intervention on commuters is 11%. One of those is a product becoming central to someone's week; the other is noise that a seasonality shift will bury. And the 167% number is why you would take the smaller segment: you can actually tell whether you succeeded, and a group whose behaviour changes that much is a group that will tell you what to build next.

Occasional users look like they win both columns, and that row is worth a sentence out loud rather than a silent omission: doubling one session a month gets you a user who is still, functionally, not a user. "Occasional" is also a residual bucket — it is defined by the absence of behaviour, so there is no shared constraint to build against, which is exactly the failure the next section warns about.

Note what the arithmetic does not do: it does not pick for you. The gap is your judgement — the honest estimate of what solving the problem unlocks — and someone with a different read of the user gets a different answer. That is fine and it is the point. The arithmetic does not replace the judgement; it makes the judgement inspectable, so the interviewer can disagree with the number rather than with your taste.

Step 2: three solutions of different ambition

The other place answers collapse is the solution list. Candidates produce three ideas that are the same idea at three sizes — "add a shortcut, add a bigger shortcut, add a customisable shortcut" — and then pick the middle one, which was never really a choice.

Generate along ambition instead. For the chosen segment and its biggest unmet need:

1. The cheap fix. Ships this quarter, no new infrastructure, no new business model. Its purpose is to establish the floor: what do we get for almost nothing?
2. The real product. The thing you would build if the need is genuine — new surface, new data, a quarter or two of work. This is the one you should usually pick, and the one you should have the most detail on.
3. The bet. Only correct if the need is bigger than we think, and it changes what the company is. Its purpose is to show you can see past the roadmap, not to be chosen.

Then say what each costs, and choose. The choosing sentence is short and it should be explicit:

"I'd build the second one. The first is cheap but it doesn't change the frequency gap, which was the whole reason I picked this segment. The third only pays off if this group is much larger than our 4% estimate, and I'd rather learn that from shipping the second one. The cost of the second is a quarter of two engineers and a new notification surface, which we'd need to defend against the ecosystem guardrail I'll come to when I talk about metrics."

Putting it together

1Scope and goal, in two sentences▾

State the assumptions and the goal you are optimising for. "Consumer mobile, mid-size company, goal is engagement rather than new revenue." One sentence of scope, one of goal, then move. If you are still clarifying at minute six, the interviewer is already worried about the deep dive.

2Segments, then one segment▾

Name two or three real segments distinguished by behaviour and constraint, not by demographic bucket. Then choose, and justify with the reach × frequency × gap arithmetic from above. Say the numbers out loud even if they're rough — a stated estimate invites correction, and a correction is a collaboration.

3The need, stated as the user would▾

For the chosen segment, three unmet needs, ranked. Phrase each as the user's problem rather than your feature: "I can't tell whether it's worth leaving now" beats "needs a live ETA widget." The feature is your answer to the need; if you skip the need you have skipped the round's subject.

4Three solutions, one choice, one cost▾

Cheap fix, real product, bet. Pick one. Name what you are giving up. This is the whole answer in miniature, and if you are running short on time, it is the part to protect.

5How you would know▾

One primary metric and one counter-metric, and the number you would need to see to call it a success. Do not do this at length here — the next lesson is entirely about doing it properly — but never end without it. An idea with no proposed measurement reads as an idea you have not stress tested.

6The objection, raised by you▾

Finish with the strongest argument against your own answer, and your response to it. "The risk is that the frequency gap is a preference, not a blocker — that these users don't want more sessions. The cheapest test of that is [X], and I'd run it before committing the quarter."

What to carry into the next lesson

You have a segment chosen for a stated reason and a solution chosen over two alternatives. Every one of those choices rests on a claim about impact, and so far the claims are unmeasured. Next: picking a metric that decomposes into something a team can move, and finding out — with arithmetic you can run — whether the experiment you just proposed could ever detect the effect you promised.