Ask people what they think of your product and you will mostly get compliments. Not out of malice — out of politeness — which is why the answer is not data. CB Insights' aggregation of startup post-mortems found the top cause of failure was neither running out of money nor competition, but "no market need," at 42%. A later update put running out of cash at 38% and lack of market need at 35%, side by side at the top.
That is why Y Combinator assigns one of its earliest Startup School lectures to how to talk to users. Eric Migicovsky, Pebble founder and former YC partner, puts it directly: the best companies are the ones where founders maintain a direct connection to users at every stage. Interviewing is founder work that cannot be outsourced.
Rob Fitzpatrick's The Mom Test is the de facto textbook here. The premise: ask only questions designed well enough that even your mother could not lie in response. She loves you, so she will say it's a great idea — and friends, and polite prospects, will do a milder version of the same. The fault lies with the question, not the respondent.
Three principles. Talk about their life, not your idea. Ask about specific facts and past behavior, not future hypotheticals or opinions — "would you use this if it existed?" is the worst question because human predictions about their own future are contaminated by optimism. And talk less. An interview extracts data; it does not sell.
Migicovsky's set: What's the hardest part about doing this? When was the last time you encountered that problem? How do you solve it today? Have you tried anything to fix it? What don't you love about the solutions you've tried?
All five are past tense and all five point at behavior. The third and fourth are decisive: someone who has never tried to solve the problem is not someone who will pay to solve it. Following surface answers with repeated "why is that?" often reaches a cause the customer had never articulated.
Idea stage: find the problem. Locate people who have it through introductions, industry events and communities, and confirm it exists through informal conversation. You do not even need to mention your product. There is no fixed number of interviews — you stop when new conversations stop producing new learning and the same patterns repeat.
Prototype stage: identify the best first customer. Now you extract numbers rather than empathy: what the problem costs, how often it occurs, and who holds the budget and approval to fix it. The segment where all three run high is your priority. Willingness to pay is not established by asking for a price — it is reverse-engineered from money and time already spent on current workarounds.
Post-launch: measure PMF and prevent churn. The common quantitative tool is the Superhuman-style survey: ask how users would feel if they could no longer use the product, and treat over 40% answering "very disappointed" as product-market fit. Call churned users and hear the reason directly. Validate new features through behavior — preorders, prepayment, opt-ins — not statements.
What investors want from an early-stage pitch is not an elaborate TAM estimate but evidence that the founder knows the customer deeply. "We've interviewed 87 prospects; 62% lose several hours a month to this problem; 14 prepaid for a paid pilot" beats any market research citation. Cost, frequency and willingness-to-pay figures pulled from interviews become the problem and traction slides directly.
There is a local trap worth naming: in Korea, the five-year survival rate for new companies is 33.8%, below the OECD average, and the government grant path makes reviewers — not customers — the gate. Business plans get tuned to evaluation criteria rather than customer problems, producing products that pass review and fail in market. The fix is simple: take the grant, but move the validation checkpoint back to customers, and treat the first paying customer, not the award announcement, as the milestone.
None of this is founder-only. Campaign messaging, landing copy and new service concepts sit on the same trap. "What do you think of this message?" returns polite agreement, and budget gets committed on the strength of it. Ask instead when they last had the problem, what they searched at that moment, and how they are solving it now — and you get real query language, real alternative competitors, and real willingness to pay in one pass. The same discipline runs through They Tried Hiring 'AI Product Builders' and Failed and Structure Your Meetings in Four Acts.
Diminishing returns matter more than a target number. When new conversations stop producing new patterns and the same answers repeat, that segment is validated for now. YC frames interviewing as continuous across all stages rather than a phase you complete.
Because predictions about one's own future behavior are contaminated by optimism and politeness. The Mom Test principle is to ask past facts instead — when the problem last occurred, what they actually tried, what they actually spent.
Do not ask for a price. Extract three numbers: what the problem costs, how frequently it occurs, and who owns the budget and approval. Reverse-engineer from current spend on workarounds, and confirm with behavior — prepayment or a paid pilot.
It is the working threshold from the Superhuman-style survey: if more than 40% of users would be "very disappointed" to lose the product, you are considered to have product-market fit. It is a widely used heuristic rather than a hard law.
To apply what you just read to your own site, start with a free audit of where things are now.
A strategist replies within 24 hours on business days.
Nobody would upload dating videos, even for $20. Then 'Me at the zoo' revealed what people actually wanted. Five lessons on choosing market signal over founding hypothesis.
The leader who accepts everything and the leader who solves everything both damage the team. Why distance is structural in management — and how to tell healthy distance from neglect.
DocSend found investors view a seed deck for 3 minutes 44 seconds on average — and lingered 80% longer on the traction slide of decks that failed. Five strategies to fill the gap.