FIELD GUIDE

How to Do Customer Discovery: From Assumption to Decision

Customer discovery is not a hunt for compliments. It is a controlled way to replace one risky assumption with evidence from people who recently experienced the problem.

This guide takes you from a testable assumption to a clear next move: keep learning, narrow the segment, reframe the problem, run a stronger behavioral test, or stop. If the term is new to you, start with what customer discovery is.

The Process at a Glance

01

Choose one assumption
02

Recruit the right people
03

Interview without pitching
04

Record the evidence
05

Choose the next test

STEP 01

Start With One Risky Assumption

Do not begin with “Is my idea good?” That question is too broad to test. Choose the belief that would damage the idea most if it were wrong—usually who has the problem, how painful it is, or what they already do about it.

ASSUMPTION TEMPLATE

We believe [specific segment] struggles with [problem] when [trigger]. Today they [work around it], which costs [time, money, or risk].

FICTIONAL B2B EXAMPLE

We believe finance leads at multi-entity marketing agencies struggle to reconcile invoices at month-end. Today they export data into spreadsheets and chase approvals in Slack, costing several hours per close.

Contradicted if: qualifying teams close quickly without manual reconciliation, a costly delay, or an active workaround.

FICTIONAL CONSUMER EXAMPLE

We believe parents of children with food allergies struggle to plan safe school lunches each Sunday. Today they cross-check labels and keep scattered notes, costing time and creating avoidable uncertainty.

Contradicted if: families reuse a stable meal list, rarely check labels, and do not see planning as a recurring problem.

Write down what would contradict the assumption before you speak to anyone. Otherwise, every polite answer will start looking like support.

STEP 02

Recruit People Who Recently Felt the Problem

Interview people who match the segment and have experienced the situation recently. Friends, generic “founders,” and people who merely like the idea are weak substitutes.

Define the role, company or life situation that qualifies.
Require a recent trigger or real encounter with the problem.
For B2B, note whether they are the user, champion, buyer, or approver.
Keep materially different segments separate when comparing answers.

COPYABLE OUTREACH

“Hi [name] — I saw [relevant signal]. I’m speaking with [segment] to understand how they handle [situation]. Would you be open to a 20-minute conversation about the last time this came up and what you did?”

Track where qualified conversations come from. If the segment is almost impossible for you to reach, that is early evidence about distribution—not merely a recruiting problem.

STEP 03

Run the Interview Without Pitching

Your job is to reconstruct real behavior, not sell the solution. Ask for specific stories, then follow the sequence of events, decisions, workarounds, and consequences.

1. Tell me about the last time this happened.
2. What triggered it, and what made it difficult?
3. Walk me through what you did next.
4. What workaround, product, or person did you use?
5. How often does this happen, and what does it cost?
6. Who else is involved in fixing or paying for it?
7. What have you already tried to change?

DON’T ASK

“Would you pay €30 per month for this?”

ASK INSTEAD

“What did you spend the last time you tried to solve this?”

This past-behavior approach follows practical guidance from Y Combinator: learn from users before pitching to them.

STEP 04

Record Evidence, Not Impressions

Write up every conversation while it is fresh. Keep conclusions tied to a participant and a concrete incident so a memorable quote cannot overpower the rest of the evidence.

FIELD	WHAT TO CAPTURE
Participant	Segment, role, and user/buyer status
Incident	The most recent specific event and its trigger
Consequence	Frequency and cost in time, money, delay, or risk
Behavior	Workaround, spending, switching attempt, or inaction
Counterevidence	Facts that weaken or contradict your assumption
Next step	Retest, narrow, prototype, pause, or stop

Compare evidence only within the same segment. Look for recurring behavior and contradictions; do not convert a few enthusiastic quotes into a fake scientific score.

STEP 05

Choose the Next Action

There is no universal interview count that validates a startup. Decide from the quality and consistency of evidence, the gaps that remain, and the cost of the next experiment.

CONTINUE INTERVIEWING

New themes are still appearing or important buyer roles remain unknown.

NARROW OR REFRAME

Pain clusters in a smaller segment, or the incidents are real but your assumed cause is wrong.

RUN A BEHAVIORAL TEST

Repeated behavior supports the problem. Test a concierge service, paid pilot, deposit, preorder, or priced landing page next.

PAUSE OR STOP

Good interviews in the exact segment reveal no recurring incident, costly consequence, workaround, or meaningful commitment.

A positive result does not mean “build everything.” It means the idea earned a stronger, more expensive test.

FICTIONAL EXAMPLE

From Broad Assumption to Narrow Test

A founder assumes that small agencies struggle with month-end invoice reconciliation. Interviews with single-entity agencies reveal annoyance but no serious workaround or budget. That is counterevidence—not a reason to improve the pitch.

The pattern changes among multi-entity agencies. Finance leads describe repeated delays, spreadsheet checks, approval bottlenecks, and several hours of manual work every month. The pain is narrower than expected, but more consequential.

NEXT DECISION

Resegment around multi-entity agencies and manually deliver one reconciliation cycle before building software. Measure access to the required data, time saved, and willingness to pay for the next cycle.

Know Your Riskiest Assumption?
Stress-Test It Before the Interviews.

Roast Rocket challenges your founder fit, market, timing, business model, and moat—free.

LAUNCH YOUR ROCKET