Customer Discovery process — 5 steps
Steve Blank's 5-step structure for testing whether a startup idea will work — before you build it. Each step has a specific output and a common pitfall.
State your hypotheses
1–2 days
Write down what you believe before talking to anyone. Who is the customer? What is their problem? Why is your solution better than alternatives? Treat as testable hypotheses, not facts.
Output:A 1-page hypothesis doc with named customer segment, named problem, and 3–5 testable claims you can falsify.
Pitfall:Vague hypotheses ("startups need productivity tools"). Force specificity: "Solo founders managing 5–15 client projects struggle with X."
Get out of the building
2–3 weeks
Steve Blank's most-quoted line. Recruit 20–30 candidates from your ICP. Cold outreach (5–10% reply rate), warm intros, paid platforms. Expect 30% of recruits to actually show up — recruit more than you think you need.
Output:A calendar with 15–20 booked interview slots over the next 4–6 weeks.
Pitfall:Recruiting only friends or warm contacts. They're polite — you'll walk away with false confidence. Mix warm with cold.
Test the problem first
3–5 weeks (concurrent with step 2)
Before showing your solution, confirm the problem is real, severe, and current. Use Mom Test–style questions: "Tell me about the last time...", "What did you do?", "What was the impact?"
Output:A confirmed problem statement (or a falsified one). Patterns emerge around interview 12–15.
Pitfall:Pitching during the interview. "Would you use a tool that..." is fatal — you contaminate the data with hypotheticals.
Test the solution second
2–3 weeks
Once the problem is confirmed, test whether your specific solution resonates. Use mockups, wireframes, smoke-test landing pages — NOT a finished product. The goal is signal, not delivery.
Output:Reactions to specific solution shape: which features land, which feel like noise, what's missing. Pre-orders or LOIs from any subset = strong signal.
Pitfall:Building too much before testing. Founders skip this step and go straight to MVP. The wireframe-to-mockup phase is what separates Discovery from Validation.
Pivot or proceed
1–2 weeks
After 20+ conversations, three outcomes: hypotheses confirmed (proceed to Validation), partially confirmed (pivot — usually customer or problem framing), falsified (kill it). The hardest move is killing fast.
Output:A signed decision: PROCEED + named next step / PIVOT + new hypothesis / KILL + postmortem.
Pitfall:Mixing partial confirmation with confirmation. "Most people liked it" ≠ "we have validated demand." Be ruthless about the difference.
Compress Step 1 into 30 minutes
The hypothesis-writing step is where most founders get stuck — vague hypotheses produce vague interviews. We built GoNoGo as the prep — a 30-min voice session that surfaces hypotheses you didn't know you had, then tests them against synthetic personas before you spend 6 weeks recruiting real interviews.
Generate hypotheses free →30 min · No credit card