How to validate a product idea in 6 steps

To validate a product idea, stop asking people whether they like it: find out whether they already spend time or money on the problem, then ask them to commit something real. Opinions are free, so they prove nothing. Behaviour costs something, so it does. The six steps below show how to do that through conversations, on one example.

A validated idea that wasn't

CatChow, an online pet-food shop, plans a subscription: the same food delivered every month, 10% cheaper than in the shop. Before the company signs contracts for packaging and couriers, Sofia, a product manager, calls 20 customers to check demand. A typical call:

Sofia: We're launching a subscription: your cat's food arrives every month and costs 10% less. Would that be useful for you?

Olivia: Oh, that sounds great! I'd probably sign up.

17 of the 20 answer like this (85%), and the team calls the idea validated. A week later Sofia sends those 17 people a pre-order link. Two of them pay.

Nobody lied. Olivia was asked to judge an idea and predict her own future, and a kind yes cost her nothing. The result is a false positive: evidence that says "go" when the honest answer is "we don't know yet".

Here is the same call with the idea kept out of it:

Sofia: When did you last run low on food for Whiskers? What happened?

Olivia: Three weeks ago. I noticed on Saturday night that the bag was nearly empty, and the shop near me doesn't stock his brand. I paid 6 dollars for express delivery on a 28-dollar bag.

Sofia: Did you change anything after that?

Olivia: I set a reminder on my phone. I also tried auto-reorder on a marketplace last year, but it sent the wrong flavour twice, so I cancelled.

No opinion about the subscription, and far more information. The problem happened recently and cost about 21% on top of the price of the bag. Olivia built a workaround, and she has already tried a similar service and dropped it for a specific reason.

Step 1. Write down the riskiest assumptions

An idea is a bundle of assumptions. List them and mark the ones that would kill the plan if they proved false. For the subscription:

Owners run out of food, or find reordering annoying, often enough to want it solved.
They are willing to commit to one food for months, although cats change their minds.
A 10% discount is a good enough reason to pay in advance.
CatChow can deliver on schedule at a cost that leaves a margin.

The first three are market risk: do customers have the problem, and will they pay to solve it? Conversations test this well. The last is product risk: can you build and run it? That takes a calculation or a pilot.

Step 2. Decide who exactly to talk to

"Cat owners" is not a segment. A student with one cat and a breeder with twelve have different problems, so their stories never repeat.

Slice the group by motive, not by age or income: who needs this most, why, and do they already pay or improvise to cope? Then name a place where you can find them. Sofia ends up with two candidates:

owners of cats on a prescription diet, found in vet clinic waiting rooms and online groups about feline kidney disease;
owners of one healthy cat with a pet shop nearby, found among CatChow's current customers.

If you can't name the place, the segment is still too broad. The lesson Customer slicing shows the method step by step.

Step 3. Ask about the past, not about your idea

The second call with Olivia worked because of three habits: the rules for questions that get honest answers. Keep your idea out of the conversation and talk about the person's life. Ask about one specific occasion that already happened, not about what they "usually" do or "would" do. And let them finish the story instead of describing your solution.

Three or four questions about the last time the problem came up, what it cost and what the person tried are enough for a 15-minute chat. For a ready list, see customer interview questions; the lesson Three rules for questions that get honest answers covers the rules themselves.

Step 4. Look for evidence that the problem is real

Sort what you hear. Words about the idea or the future are weak; facts about what the person has already done are strong.

Signal	What it really tells you	Evidence
"Sounds great, I love it"	The person is polite	None
"I'd definitely buy it"	A guess about the future, made to please you	None
"It's annoying, but I've never tried to fix it"	An inconvenience they live with	Negative
A recent story with details: day, place, price	The problem exists and happened not long ago	Medium
A workaround: reminders, a spare bag, a spreadsheet	They already put effort into coping	Strong
Money already spent: express delivery, a pricier shop	The problem has a price, and a benchmark for yours	Strong
They searched for a solution or tried one	They are a buyer, not only a complainer	Strong
They give up time, reputation or money for your solution	Real demand from this person	Strongest

Step 5. Ask for a commitment and read the answer

Once you understand the problem, show the idea. Compliments will come back, so don't ask for an opinion. Ask for something that costs the person:

Time: use a trial box for two weeks and note what went wrong.
Reputation: introduce you to the administrator of the owners' group or to their vet.
Money: pre-order the first box for 30 dollars.

Then read the answer. A yes with a date or a payment is a signal. "Keep me posted" is a polite stall: ask what is in the way. A clear no is a result too: it shows what blocks the purchase. The lesson Commitment currencies explains how much each kind of commitment is worth.

Step 6. Decide

After two weeks Sofia has 12 conversations in her notes:

Segment	Conversations	Recent run-out story	Workaround or extra spending	Pre-ordered
Cat on a prescription diet	6	5	5	3
One healthy cat, shop nearby	6	1	0	0

Taken together, 3 of 12 people pre-ordered, a lukewarm 25%. Split by segment, it is 3 of 6 against 0 of 6. The 90 dollars collected is pocket money, but it was paid for a product that doesn't exist yet.

These numbers are not statistics: six conversations can't estimate a conversion rate. They show where the problem lives, which is enough to choose one of four decisions:

What you found	Decision
The problem is real and people commit	Continue, and move to a quantitative check
The problem is real for one group, the rest shrug	Change the segment: narrow down to those who commit
People keep describing a different, more painful problem	Change the problem
No workarounds, no spending, no commitments in any segment	Stop, and keep the budget

CatChow changes the segment: the first version of the subscription is for owners of cats on prescription diets.

How many conversations are enough

Continue until you stop hearing anything new. With a well-defined segment that often happens after three to five conversations. Count per segment, not in total.

If ten conversations give you ten different stories, more of them won't help. The segment is too broad, so go back to step 2.

Where interviews end and numbers begin

Conversations tell you whether the problem is real, for whom and why. They can't tell you how many people will buy. That takes a quantitative check: a landing page that takes pre-orders, and later an A/B test of the price or the offer.

Suppose 800 owners from clinic groups see CatChow's pre-order page and 24 of them pay: 3.0%. Even with 800 visitors, the true rate could plausibly be anywhere from about 2% to 4%. How much traffic a reliable answer needs is covered in A/B test sample size.

Common mistakes

Pitching in the first sentence. After that, every answer is about your idea, and polite.
Counting smiles. "12 conversations, 12 positive" reports the mood. Count commitments.
Asking more people the same question. 200 polite yeses mislead as much as 20.
Skipping the ask. Avoiding a no leaves you guessing for months.

What to do this week

Write down three assumptions that would kill the idea.
Pick one narrow segment and the place where you will find it.
Have five conversations about the last time the problem came up. End each with "Who else should I talk to?"
Prepare one request for each currency: time, reputation, money.
Count commitments, not compliments, and decide.

Key takeaways

Opinions and promises are not evidence; past behaviour, workarounds and money already spent are.
Test the riskiest assumptions first, with one narrow segment you know where to find.
Ask for time, reputation or money, and treat a clear no as a result.
Interviews show whether and why; landing pages and A/B tests show how many.

FAQ

How do I validate a product idea without a product?

Talk to people in one narrow segment about how they deal with the problem today. If they already spend time or money on it, describe the solution and ask for a pre-order, a trial or an introduction. None of that needs a finished product.

How many customer interviews do I need to validate an idea?

Often three to five per well-defined segment, until the stories start repeating. If every conversation is different, narrow the segment.

Is a survey enough to validate an idea?

A survey collects opinions and predictions, so it has the same flaw as "do you like my idea?". It is useful later, to measure how common an already understood problem is.

Learn it hands-on

The free lesson Why "do you like my idea?" is useless takes Sofia's first calls apart line by line, with quizzes and a practice task checked by AI. It is part of the free course Customer Development: Talking to Customers.

How to validate a product idea before you build it