Problem discovery checklist: 12 points to review in 2027

A checklist is only useful if its items can fail. Most discovery checklists are written so that reading them counts as passing, which is why people tick them and learn nothing.

These twelve have failing states attached. Work through them in order: the early ones are cheap and the later ones depend on them. If an item fails, that is the next piece of work, and none of it requires you to build anything.

What to take away

• Every item here has a stated failing condition. If you cannot fail it, you have not checked it.
• The first four cost nothing and remove most of the wasted months.
• Reaching the end is not permission to build. It means you know enough to put something specific in front of someone.

Before you talk to anyone

1. The problem is written without your solution in it. Read your statement and delete every clause that describes a missing product. If nothing survives, you have a product description with the tense changed.

Fails if the sentence only makes sense to someone who already knows what you plan to build.

2. The statement names five things. Who exactly, in what situation, doing what today, at what cost, how often. Each is a separate claim that can turn out false.

Fails if any part is a category rather than a description, especially the who.

3. You can name a route to twenty of these people that does not depend on who you know. Write down where they gather.

Fails if every route runs through a personal introduction. Those are still worth using, but you have a reach problem you have not costed.

4. You have written what would disappoint you. One sentence, before any conversation.

Fails if no answer anyone could give would count as bad news.

While you are finding out

5. You are asking about last time, not next time. Every question aims at a specific past event.

Fails if your notes are full of would, could and probably. See the questions that recover the past, and the guidance on writing survey questions for how much a question's shape decides the answer you get back.

6. You have found what people do today, in detail. The spreadsheet, the junior colleague, the agency, the workaround, the doing nothing.

Fails if you cannot describe the current method well enough to perform it yourself.

7. You know what the current method costs. Hours, money, rework, risk, or a relationship, with a figure the person gave you rather than one you inferred.

Fails if the cost only appears in your own summary of the conversation.

8. Someone is already spending to reduce this. Paying, hiring, building, or working weekends.

Fails if nobody has ever spent anything on it. That is not disqualifying, but it moves the burden of proof onto you. Existing spend is also the evidence the SBA's guide to market research and competitive analysis asks for before a market is treated as real.

Before you conclude anything

9. You have talked to the person who pays, not only the person who suffers. These separate as soon as the buyer is an organization, and often before.

Fails if everyone you spoke to would have to ask somebody else for the money.

10. You have found the trigger. The event that makes someone look for a change: a failure, a hire, an audit, a renewal, a threshold, a rule change.

Fails if the honest answer is that people would switch when they get round to it. They will not.

11. Your sample is not just the people who were easy to reach. Count how many came through your own network.

Fails if the answer is most of them. Two columns in your notes fix this: what you learned about the problem, and what you learned about reach. Market validation is where the second column gets settled.

12. New conversations have stopped surprising you. You can predict what someone will say about their current method before they say it.

Fails if every conversation still moves the statement, which means keep going, or if you stopped at three, which means you stopped because they agreed with you.

The one item people skip

Nine. It gets skipped because the person who feels the problem is the pleasant conversation and the person who releases money is the awkward one.

The cost of skipping it does not appear until you have something to sell, at which point the enthusiasm you collected turns out to belong to people who cannot buy. Move it earlier than feels natural: in the first conversations, ask who signed off the last thing they bought like this, and what that took.

What passing means

Twelve passes is not a verdict on the idea. It means the description of the problem is now specific enough that an offer can be wrong in a useful way, which is the only state from which the next test tells you anything. How the pieces connect is set out in problem discovery; whether you can afford the answer is a separate question, covered in founder readiness.

Common questions

How long should this take?

The first four are an afternoon. The middle four take as long as real conversations take, usually weeks rather than months, because scheduling is the constraint. The last four are a review, not new work.

Can I do the items out of order?

You can, but items seven and eight are much harder to answer if you skipped five, because you will have opinions instead of accounts. Item nine can and should start earlier than its position suggests.

What if the problem changes halfway through?

That is the checklist working. Rewrite the statement, date the new version, and keep the old one. The sequence of statements is the record of what you learned, and a statement that never moved usually means nobody looked hard.

Does this replace competitor research?

No. This establishes that a problem exists and is worth money to someone. What else is offered against it, and why people stay with what they have, is a separate reading: see competitor validation.