Developed by Steve Blank (The Four Steps to the Epiphany, 2005), with later refinements by Eric Ries (2011) and Rob Fitzpatrick (2013).
Tests whether a problem is real and who has it through many interviews about past behavior, patterns across users, and requests for commitment before anything is built.
Four claims from a scheduling plan, judged against twelve interviews
The plan rests on four claims, and twelve coordinators walked through their last call-out. For each claim, read what the interviews turned up and judge it: confirmed, refuted, or not yet tested. The reply says what the evidence supports and what the founders do next; the bar counts the claims settled by real evidence. Illustrative.
Founders
The interviews
Founders
The interviews
Verdicts the evidence supports: 0 of 4Verdicts the evidence supports: 1 of 4Verdicts the evidence supports: 2 of 4Verdicts the evidence supports: 3 of 4Verdicts the evidence supports: 4 of 4
Claims settled by real evidence0of 4
- The plan (Founders). The business plan for home-care scheduling software rests on four claims. The founders booked twelve coordinators they had never met, at agencies of 50 to 200 caregivers that schedule in spreadsheets, and asked each to walk through the last call-out.
- Hours, counted (The interviews). Walked through step by step, the coordinators' time spent rebuilding the schedule after call-outs comes to about three hours a week. None comes near ten.
- Judge the first claim (Founders). The plan says coordinators lose ten hours a week rebuilding the schedule after call-outs. The walk-throughs put it at about three.
- Confirmed The time is real, but the claim named ten hours and the evidence shows about three. The founders drop the ten-hour figure and stop pitching time saved. Stretched to fit: a claim that survives any number tests nothing.
- Refuted Refuted. Twelve walk-throughs put the loss at about three hours a week, a third of the claim. The founders drop the ten-hour figure and stop pitching time saved. Right: the number could fail, and it did.
- Not yet tested It was tested: twelve accounts of real call-outs, and the hours came to about three. The founders drop the ten-hour figure and stop pitching time saved. Too cautious: past behavior is the test, and it has answered.
- The workaround (The interviews). Eight coordinators, at agencies with no tie to each other, describe the same workaround: a group text to every caregiver, the shift going to whoever answers first, qualified or not. Three of them lost a client over a shift filled by the wrong person.
- Judge the second claim (Founders). The plan says coordinators have no reliable way to fill a call-out. Eight of twelve, unconnected, describe the same group text, and three lost a client to it.
- Confirmed Confirmed. The same behavior from eight independent agencies, with lost clients as its cost, is a pattern. The founders make it the lead claim: fill a call-out with a qualified caregiver within two hours. Right: a repeated behavior with a cost attached.
- Refuted They do fill shifts, but by texting everyone and taking whoever answers, and three lost a client that way. That is the claim confirmed, with its sharpest form: qualified, within two hours. Misread: a workaround that fails is evidence of the problem.
- Not yet tested Eight unconnected accounts of the same workaround, and three lost clients, are real evidence. The founders make filling a call-out with a qualified caregiver within two hours the lead claim. Too cautious: eight independent accounts are a pattern.
- The praise (The interviews). Nine of the twelve say a better scheduling tool would be great. Asked about the past year, none has looked for one, tried one, or set money aside for one; the spreadsheet is still what they use.
- Judge the third claim (Founders). The plan says coordinators want a full scheduling product in place of their spreadsheets. Nine say a better tool would be great; none has acted on it.
- Confirmed Nine compliments about a product that does not exist are opinions, and no one has looked, tried, or paid. The full product stays out of the pilot until a commitment tests it. Politeness counted as evidence.
- Refuted Nothing here argues against it either: no one was asked to act, so no one declined. The full product stays out of the pilot and waits for a test that asks for a commitment. Too far: an untested claim is not a failed one.
- Not yet tested Not yet tested. Praise for a product that does not exist is an opinion, and no one has acted on it. The full product stays out of the pilot and waits for a test that asks for a commitment. Right: opinions leave the claim where it started.
- The ask (The interviews). At the end of each interview the founders ask for a paid pilot of call-out filling alone at $300 a month. Four of the twelve agencies sign; the rest decline or want to wait.
- Judge the fourth claim (Founders). The plan says agencies will pay $300 a month for software that fixes call-outs. Asked for exactly that, four of twelve signed.
- Confirmed Confirmed, for call-out filling. Four signed commitments are the evidence politeness never supplies. The founders run the pilot with four agencies and count shifts filled by a qualified caregiver within two hours. Right: money committed is the strongest answer.
- Refuted Eight said no or not yet, but four signed at the full price, which is the claim holding for a first set of buyers. The founders run the pilot with those four and count shifts filled within two hours. Misread: the claim never said every agency.
- Not yet tested A signed pilot at the stated price is the test, and four agencies passed it. The founders run the pilot with those four and count shifts filled by a qualified caregiver within two hours. Too cautious: a commitment is real evidence.
An example
Two founders planned scheduling software for home-care agencies. The plan rested on four claims: coordinators lost ten hours a week rebuilding the schedule after call-outs; they had no reliable way to fill a call-out; they wanted a full product in place of their spreadsheets; and agencies would pay $300 a month to fix call-outs. Before writing code they booked twelve coordinators they had never met, at agencies of 50 to 200 caregivers on spreadsheets.
Each interview walked through the last call-out. Rebuilding took about three hours a week, so the first claim fell. Eight of the twelve, at unconnected agencies, described the same workaround: a group text to every caregiver, the shift going to whoever answered first, qualified or not. Three had lost a client that way. Nine said a better tool would be great, yet none had looked for one, so the third claim stayed untested. Asked for a paid pilot at $300 a month, four agencies signed. The founders dropped the ten-hour figure and made filling a call-out with a qualified caregiver within two hours the lead claim.
Why it works
A product idea is a stack of guesses about who has a problem, how much it hurts, and what they do about it now. Customer discovery writes those guesses down as claims that can fail, then tests them against what people have actually done, before the cost of being wrong includes a built product.
Two rules make the test honest. Interviews ask about past behavior, so a coordinator's description of last Tuesday counts and her opinion of a hypothetical product does not. And one story is an anecdote: a claim earns belief only when several people who do not know each other describe the same behavior, here the group text. The last check is a commitment, such as a paid pilot, because money is the one answer that politeness never supplies.
The thinking behind it
Steve Blank Steve Blank, a serial entrepreneur and teacher of entrepreneurship, set out customer development in The Four Steps to the Epiphany (2005), first self-published out of his practice and teaching. A new venture, he argued, should treat its business model as untested hypotheses and test them with customers outside the building. Customer discovery is the first of its steps, where the founders learn whether the problem and the customer are real.
Eric Ries Eric Ries joined Blank's customer development to ideas from lean manufacturing in The Lean Startup (2011). His method builds a minimum viable product and runs it through a build, measure, learn loop, built around validated learning. For discovery, that sharpens the test: a claim moves forward only when evidence from customers supports it, however much the team likes it. The book is an entrepreneur's synthesis, written for practitioners.
Rob Fitzpatrick Rob Fitzpatrick's The Mom Test (2013) supplied the interview discipline that discovery depends on. Its premise is that even a founder's own mother will give a kind answer about the idea, so the questions stay on the customer's life and past behavior and never touch the idea itself. That rule is what makes a round of interviews a test: an account of what a coordinator did at the last call-out can contradict a claim, and an opinion about a product that does not exist cannot. It is a short practitioner's guide.
How to run it
- Write the claim so it can fail. Name who has the problem, when it happens, and how much it costs, with a number that could turn out wrong.
- Define the segment by behavior. Describe the customers by what they do and the setting they work in, and leave demographics for later.
- Interview ten or more strangers. Ask each to walk through the last time the problem came up, and avoid friends and warm contacts.
- Log behavior, ignore praise. Write up what each person did and spent the same day.
- Look for patterns across independent people. Treat a behavior as real only when three or more unconnected people describe it.
- Ask for a commitment. Ask for a pilot, a deposit, or an introduction, and count who says yes.
- Rewrite, narrow, or drop the claim. Change the claim to fit the evidence and test the new one.
Where it goes wrong
Founders often interview the wrong people. Friends, former colleagues, and enthusiastic early followers agree for social reasons, and the patterns they produce are patterns of politeness. Recruit strangers who fit the segment.
A second failure is keeping the claim alive by stretching it. When the interviews contradict the ten-hour claim, it is tempting to reword it until it fits everything, which leaves it testing nothing. If no number in the claim can be wrong, rewrite it.
The third is stopping at enthusiasm. Twelve warm interviews with no pilot, deposit, or introduction mean the problem is real and still not urgent enough to pay for. Count commitments before building.
When to use something else
- Mom Test. When the work is the single conversation and the need is questions that keep politeness from passing for evidence.
- Jobs to Be Done. When the problem is confirmed and the question is what progress customers are trying to make and what they would switch from.
- Randomized Controlled Trial. When a product exists and a change can be tested on live users with a control group.
Where it sits in SOLVE eX
Go to the source
- The Four Steps to the Epiphany, 2005
- The Lean Startup, 2011
- The Mom Test, 2013