How Customer Discovery Actually Works

Customer discovery is not market research and it is not validation. It is the work of trying to prove yourself wrong while it is still cheap to be wrong. Founders who do it well are not the ones who talk to the most people; they are the ones who ask questions that could have come back with an answer they did not want, and who change something when they do. It also never really ends. As long as you run the company you will be doing some version of it, so it is worth getting good at.

The process, in order

1. 1

   Decide what would change your mind

   Before any conversation, write down the belief your company depends on and what evidence would kill it. That is what turns a pleasant chat into problem validation: you are not collecting encouragement, you are running a test with a result you have agreed to accept in advance.

   Most founders skip this and end up gathering evidence for a verdict they already reached. If you cannot name the finding that would make you stop, you are not doing discovery, you are doing reassurance.

2. 2

   Pick who to talk to, precisely

   Write down who you are looking for tightly enough that two people could sort the same company the same way. That is your ideal customer profile, and at this stage it is a hypothesis rather than a finding. Be clear whether you are describing a company, a role, or a person, because a customer segment vs persona mix-up is what produces twenty interviews that cannot be compared with each other.

   Narrow beats representative. Ten conversations with people who share a situation will teach you more than fifty spread across a market, because only the first set lets you tell a pattern from a coincidence.

   Do this in Lev

   Lev drafts this profile from your idea and checks it against real numbers: how many people match, and which choices make the list too broad to learn from.

   Build your ICP with Lev
3. 3

   Get the conversations by hand, then add a little scale

   Nobody will book the meeting for you. Recruiting early users is manual work, and early on every conversation comes from a personal ask: your own network, one introduction at a time, a cold message that proves you understand their day. This stage feels like it does not scale, and that is the point, because the thing you are buying is not volume but access to people who will tell you the truth.

   Set the expectation plainly when you ask. You want twenty minutes about how they currently handle something, not a demo and not a favor. People agree to that far more often than founders expect, and it produces a different conversation than a pitch does. Meet in person when you can, and do not pay people with gift cards. The reward that works is solving a real problem of theirs, even one unrelated to your product, which is also what earns the next introduction.

   End every conversation by asking who else you should talk to, and make saying yes effortless: offer to send a two-line note they can forward, already written. One good meeting that turns into three referrals is how a list of twenty becomes fifty without buying any of it.

   Once the conversations are flowing, a short form adds scale without replacing them. Use it three ways in turn: as your interview script while you refine the questions, as a few questions sent ahead of each meeting so the call starts deeper, and finally posted where your customers already gather, so the right people find you. The same handful of questions does all three jobs.

   • Background: who they are and the situation they are in, in the terms your profile uses.
   • Problems, ranked: which matter most, so you can see whether yours makes their top three.
   • Quantified value: what the problem costs them, and what they could and would pay to fix it.
   • Other solutions: what they have tried, bought or built, which is your real competition.
   • Other new products: which recent tools they have adopted, the best single sign of an early adopter.
   • Where they discover things: the people, newsletters, communities and events they actually follow.
   Do this in Lev

   The asking stays personal; the list does not have to be. Lev finds twenty people who match your profile, looks up how to reach each one, lets you mark the ones you can reach through someone you know, and drafts a first message for you to make your own.

   Find your first twenty conversations with Lev
4. 4

   Run the interview so politeness cannot ruin it

   The whole craft of a customer discovery interview is asking about the past rather than the future. What someone did last month is a fact; what they would do is a compliment. Ask what they last tried, what it cost them, what they had to work around, and what happened after.

   Guard hard against leading questions. Asking whether they would use something that fixed this gets a yes from almost everyone and tells you nothing, because agreeing is the cheapest way for a person to be kind to you. The strongest evidence you can leave with is not enthusiasm, it is a story about something they already did.

   Walk in with the order of the conversation already decided. Confirm the person matches your profile, then test the problem, whether they could pay, which feature matters most, and a real price, one step at a time, and only ask someone to become an early customer once every step has held. Each no has its own follow-up question, and the follow-up is often worth more than the yes would have been. When the conversation is going well, the money questions are not optional: always ask whether they could pay, whether they would, and how much. Even a conversation that stops at the first step can end with a referral to someone who does have the problem.

   • Ask about the last time, not about whether they usually.
   • Follow the effort, whatever they built, bought, or hacked around it.
   • Let silence run, the useful sentence is often the one after the pause.
   • Never describe your solution until the problem conversation is finished.
5. 5

   Separate what you heard from what it means

   Interviews produce quotes; decisions need patterns. Work out what people were actually trying to get done, which is the discipline jobs to be done exists to enforce, and resist rewriting their goal into the one your product serves.

   Then sort ruthlessly. A painkiller vs vitamin judgment is the one that predicts whether anyone pays: a painkiller has a cost attached that people can name, while a vitamin gets sincere praise and no budget. Both sound positive in a transcript, which is exactly why this has to be a deliberate step rather than an impression. A simple scale helps: hair on fire, painful, annoying, or nice to solve. The first two are painkillers; the last two are vitamins.

   Put a number on it so conversations can be compared. Ask each person to rate how much a solution would be worth to them from one to ten, how they solve it today, and what that costs. Then check the rating against what they have actually spent. Most people who say they need something have spent no real time or money looking for it, and that gap is a warning about them as early adopters and sometimes about the idea itself.

6. 6

   Find the person who cannot wait

   Somewhere in the pattern is a smaller group whose problem is bad enough that they have already tried to solve it themselves. That is your early adopter, and the evidence is behavioral: a spreadsheet, a contractor, a workaround they are embarrassed by.

   Score for it rather than guessing. The people who have recently adopted other new products, and who follow the sources where new products surface, are far more likely to take a chance on yours. The form from stage 3 gives you that score for everyone who fills it in, before you have spoken to them.

   From that group, choose an ideal first customer deliberately rather than taking whoever is friendliest. The first customer shapes the product more than any other, so the right one is the one whose problem is most representative of the market you want, not the one who replies fastest. Aim for three to five who pay. That is the group whose behavior, not opinion, will tell you what the next customers have in common.

7. 7

   Turn a believer into a commitment, then into a teammate

   Interest is free, so convert it into something that costs the other person a little. A design partner arrangement is the usual shape: their team's time on a regular cadence and access to the real workflow, in exchange for influence and favorable terms. The act of agreeing to specifics is itself the test.

   Be honest with yourself about what a pilot vs paid contract tells you. A free pilot measures curiosity; money, a signed order, or a named internal budget measures priority. You do not need revenue at this stage, but you do need to know which of the two you are holding.

   Then build them into how you work. Plan, ship and collect their feedback on every cycle, every two weeks or so, and bring the rest of the team to some of those conversations so everyone building the product has met the person it is for. Add new early customers as you go, and be willing to let go of one who keeps pulling the product away from the profile. A bad-fit early customer costs more than no customer, because their requests look exactly like evidence.

8. 8

   Keep a rhythm

   Discovery produces more than anyone can hold in their head, so give it a home from the first week. Keep it in one sortable list: one row per person or company, a column for each attribute in your profile, and the answers from stage 3. Scoring that list against your profile, and checking the score against who actually buys, is its own discipline, covered in How Picking an ICP Actually Works. What belongs here is the rhythm that turns the list into decisions.

   Then put it on a calendar, because a cadence is what turns a pile of conversations into decisions. It works like a weekly business review for customers, with a longer look each month and each quarter.

   • Weekly: the new conversations, what was discussed, and the one or two things you now believe that you did not last week.
   • Monthly: the numbers, what worked and what did not, whether the proposition and the pitch still hold, next month's targets, and whether the scoring still predicts.
   • Quarterly: the whole picture again, the customer, the value, the feature set and the business model, along with the broad go-to-market approach, pricing and the competition.
   Do this in Lev

   Lev keeps the list for you: everyone it found or you added, on one board from found to contacted, replied, discovery call and interviewed, with whether you can reach each person warm or cold. The weekly look starts from the board instead of from your inbox.

   Keep score in Lev
9. 9

   Decide whether you have fit, or another lap

   Discovery ends when behavior replaces opinion. Product-market fit shows up as people returning without being prompted, getting annoyed when the product breaks, and telling colleagues unasked. The cleanest read is cohort retention: if a curve flattens rather than decaying to zero, something is holding.

   It helps to know where you are on the way. Most founders pass the same markers in roughly the same order, and naming the next one is more useful than asking whether you have fit yet.

   The honest outcome is usually another lap. Discovery is not a phase you complete before building; it is the loop you stay in, running tighter and tighter as the questions get more specific. Fall in love with the customer and with their problem first. The love for your solution follows, and it is better founded for coming second.

   • A clear, quantified ideal customer profile you can run against a list.
   • More than twenty discovery conversations with people who match it.
   • A prioritized pipeline, sorted by a lead score you trust.
   • At least three paying early adopters.
   • More than twenty paying customers, or around $10,000 in monthly recurring revenue: the first sign of repeatability.
   • Unit economics that hold, on the way to predictable revenue.