Customer discovery is the work of finding out whether a problem is real, who has it, how often it bites and what they do about it today, before you build anything. The customer discovery process runs in five steps: write down your riskiest assumptions, decide exactly who to talk to, recruit people who match, interview them about what they have actually done, and decide what the evidence tells you.
This guide covers each step, the customer discovery questions that produce evidence and the ones that produce flattery, how many interviews you need, what good evidence looks like, and a free interview guide you can copy.
What is customer discovery?
The term comes from Steve Blank’s customer development model, where customer discovery is the first of four stages, ahead of customer validation, customer creation and company building. The idea is simple: before you build, get out and test whether the problem you are solving exists in the form you think it does.
In practice that means talking to people, mostly in one-to-one interviews, about their lives rather than your idea. Rob Fitzpatrick’s The Mom Test puts the rule well: ask about what people have done, not what they would do, because opinions about the future are free to give and cost nothing to get wrong.
Customer discovery vs customer validation
They are often mixed up, and the mix-up is expensive. Discovery asks whether the problem is real and worth solving. Validation asks whether your solution is one people will adopt. Discovery comes first, and you can do it without anything to show. Validation needs a concept, a prototype or a price to test.
Skipping discovery and going straight to “would you use this?” is how teams end up with a polished answer to a problem nobody was trying to fix. Once discovery says the problem is real, product validation is the next step.
The customer discovery process in five steps
1. Write down your riskiest assumptions
Every idea rests on a handful of beliefs: that a certain group has a certain problem, that it happens often, that it costs them something, that they are not happy with what they use now. Write each one down and ask of it: if this is wrong, does the idea die? Rank them by that, and design your interviews around the top three. Discovery that tests nothing in particular tends to confirm whatever you hoped.
2. Decide exactly who to talk to
“Small business owners” is not a target. “Owners of cleaning businesses with three to ten staff who quote jobs by hand” is. Define the people who should have the problem, then screen on behaviour rather than job title: someone who dealt with the problem in the last month is worth five people who say it sounds familiar.
3. Recruit people who are not your friends
Friends, family and your own network are the easiest people to reach and the worst to learn from, because they want to be supportive. Recruit strangers who match your screener, offer a fair incentive for their time, and expect to screen out a good share of applicants. Recruitment is usually the slowest part of discovery, and the part most likely to be quietly compromised to save time.
If your audience is hard to reach, such as senior B2B buyers or clinicians, we recruit verified professionals and run the interviews for you, from screener to findings.
4. Interview about the past, not the future
Thirty to forty-five minutes, one person at a time, with a discussion guide but not a script. Spend most of it on the last time the problem happened: what triggered it, what they did, what it cost. Keep your idea out of the conversation until the end, or leave it out altogether. The moment you pitch, people start reacting to you instead of telling you about their world. Our post on customer interview mistakes covers the traps in detail.
5. Synthesise by counting, then decide
After the interviews, do not reach for your favourite quotes. Go back to your assumptions and count: how many people had the problem recently, how many had built a workaround, how many could put a cost on it. Then make a call. Carry on if the evidence holds, change the target group or the problem if it is pointing somewhere else, or stop if it is not there. A discovery round that could not have told you to stop was not really testing anything.
Customer discovery questions: an interview guide you can copy
The questions that produce evidence are open, anchored in something that already happened, and leave your idea out. This customer discovery template puts them in order as a 40-minute interview built around one recent event, with a note under each question on why it is there. Swap the bracketed parts for your problem, keep the order, and resist adding questions about your idea.
Customer discovery interview guide
- Tell me about your role and what a typical week looks like. Context first: you need to know where the problem would sit in their week before you ask about it.
- Which parts of that take the most time or cause the most hassle? Listen for whether the problem comes up without prompting. If it does, that is a good sign.
- When did you last deal with [the problem]? Ask for the last time, not in general. A specific recent story is evidence, a general opinion is not.
- Walk me through it from the start. What triggered it? Let them tell it. The detail they include, and the parts they skip, show you what matters to them.
- What did you do about it? The single most important question. If the answer is nothing, the problem may not be worth solving.
- How did it turn out? Listen for frustration, effort, and anything they would do differently next time.
- Have you tried anything else to fix this? What happened? Past attempts show how hard they have looked for a solution, and why the existing options fell short.
- What does it cost you when it happens, in time, money or hassle? Ask for a rough number. Problems with no cost attached rarely get budget.
- How many times has this happened in the last month? A count, not “usually”. Frequency separates an annoyance from a priority.
- Who else is involved, and who would decide on a fix? In a business, the person with the problem is often not the person who would buy the fix.
- Where does this sit against everything else on your plate? A real problem that ranks tenth will not get acted on. You want the ones near the top.
- If it were solved, what would be different for you? Ask late. It shows the outcome they want, not the feature to build.
- Is there anything I should have asked but did not? People often save the most useful point for the end, once they know what you are interested in.
- Who else should I talk to about this? The cheapest recruitment you will get, and a test of whether they care enough to introduce you.
- Within an hour, note the specific story, the workaround, the cost and the frequency.
- Mark each of your top assumptions as supported, contradicted or untested by this interview.
- Write down one thing you did not expect. If you have nothing for three interviews running, you are near saturation.
Want customer discovery done for you?
We write the screener, recruit LinkedIn-verified people who match it, moderate every interview with your team setting the guide and observing, and come back with what the evidence says about each assumption.
Questions to avoid, and what to ask instead
| Avoid | Ask instead |
|---|---|
| Would you use a tool that does X? | When did you last deal with this? What did you do? |
| How much would you pay for this? | What do you spend on solving it today? |
| Don’t you find it frustrating when…? | How do you feel about the way you handle it now? |
| Do you think this is a good idea? | What have you tried before? Why did you stop? |
| How often do you usually…? | How many times did it happen last month? |
The pattern is the same in every row: replace a hypothetical or a leading question with one about a specific thing that already happened. If you want to put a price on the problem later, that is a separate study: see how to run a pricing survey.
What good evidence looks like
Not every positive answer counts. Some signals are worth a lot, some are close to worthless, and the worthless ones are the ones that feel best in the room.
| Weak signal | Strong signal |
|---|---|
| “That sounds interesting.” | They describe a specific time it happened in the last few weeks. |
| “I would definitely use that.” | They have built a workaround: a spreadsheet, a script, a manual routine. |
| “Let me know when it is ready.” | They already spend money or real time on the problem. |
| General agreement that the problem exists. | They offer to introduce you to a colleague, or ask to see it first. |
The strongest single signal is a workaround. People only build one for a problem that hurts enough to act on, and the workaround tells you exactly what your product has to beat.
How many customer discovery interviews do you need?
Usually eight to fifteen per segment. By then most rounds stop producing anything new, the point researchers call saturation. If you are talking to two very different groups, such as the people who use a product and the people who pay for it, treat them as two segments and run a set for each. Stop when three or four interviews in a row teach you nothing new, not when you hit a number.
A customer discovery example
An illustrative case. A founder believed that small agencies lose hours every week chasing clients for feedback on design work. The riskiest assumption was not that this happened, but that agencies saw it as a problem worth paying to fix.
Twelve interviews with agency owners told a split story. Ten had chased feedback in the last fortnight, so the problem was real. But only three had any kind of workaround, and most described the chasing as part of the job rather than something broken. The three with workarounds all ran agencies with more than fifteen staff, where the chasing had been handed to a dedicated account manager, which meant it had a salary attached.
The decision was to narrow the target from small agencies to agencies large enough to employ someone whose time the product would free up. Same problem, different customer, and a much clearer case for paying. That is what good discovery does: it rarely says yes or no, it tells you who.
Frequently asked questions
What is customer discovery?
Customer discovery is the first stage of building a product: finding out whether a problem is real, who has it, how often, and what they do about it today, before you build a solution. It is done mostly through interviews about what people have actually done, not what they say they would do.
What is the difference between customer discovery and customer validation?
Discovery tests whether the problem is real and worth solving. Validation tests whether your particular solution is one people will adopt. Discovery comes first and is mostly interviews about the problem. Validation comes second and tests a concept, a prototype or a price against real buyers.
How long does customer discovery take?
For one clear segment, two to four weeks: a few days to write assumptions and a screener, one to two weeks to recruit and run eight to fifteen interviews, and a few days to synthesise. Narrow professional audiences take longer, mainly because recruiting them takes longer.
Can you do customer discovery with a survey?
Not on its own. A survey can only measure the answers you thought to offer, and discovery is about finding what you did not know to ask. Run interviews first to find the problems and the language people use, then use a survey to size how widely the pattern holds.
How is customer discovery different from market research?
Market research describes a market: its size, its players and its trends. Customer discovery tests a specific assumption about a specific group of people, and is designed so that it can tell you to stop. You can have a large market and still find that the problem you planned to solve is not one anybody is trying to fix.
Talk to the right people first
Customer discovery is only as good as the people you talk to. The hardest part is reaching people who really have the problem and hearing what they mean rather than what is polite, which is where most founder-led discovery goes wrong.
We run discovery interviews end to end: screener, recruitment, moderation and a findings readout, as user research or as the first stage of product validation. To run interviews at scale yourself, the Kosmo platform asks neutral, behaviour-focused follow-up questions automatically.
Not sure which fits? Book a free 30-minute call.