Last reviewed:
Most startups don't fail from bad execution. They fail from building something nobody wanted. Validation is the discipline of finding out cheaply. This workshop is a two-week sprint with seven gates; every gate produces a written artifact you'll defend to your team and your future self.
- 1
Write the hypothesis
Use the format: 'We believe [audience] has [problem] severe enough to pay [price] for [solution] delivered via [channel].' One sentence. Every term concrete.
- 2
Enumerate falsifiable assumptions
Break the hypothesis into atoms: 'audience exists at scale X', 'problem occurs weekly', 'current alternatives cost Y', 'channel converts at Z%'. Aim for 12-20 assumptions.
- 3
Prioritize by leverage
Score each assumption by how much it would change the plan if wrong (1-5) and how uncertain you are (1-5). Test the top 3 by product first.
- 4
Pick the cheapest test per assumption
Demand: smoke test page + paid traffic. Problem severity: 10 customer interviews using The Mom Test. Willingness to pay: pre-order with refund or paid pilot. Channel: a 7-day micro-campaign.
- 5
Set falsification thresholds in advance
Write the numbers that would kill or confirm the assumption BEFORE running the test. '<2% landing page conversion = kill.' Otherwise you'll rationalize whatever happens.
- 6
Execute in a two-week sprint
One owner per test. Daily 15-minute standup. Raw data goes in a shared doc. No new tests added mid-sprint.
- 7
Decide: persevere, pivot, or kill
At end of sprint, compare results to your thresholds. Write the decision and the reason in one paragraph. If pivot: re-write the hypothesis and run the next sprint.
Startup Validation Workshop - FAQ
- How is validation different from market research?
- Market research describes what exists. Validation tests whether your specific offer changes what people will do. The currency is behavior, not opinion.
- Do I need a product to validate?
- No. A landing page, a slide deck, a Figma prototype, or a manual concierge service is usually enough. Building the product is the most expensive way to test demand.
- How many customers should I interview?
- Five interviews surface 80% of recurring patterns. Ten gives confidence. Twenty starts to repeat. Do them in batches of five so you can adjust the script.
- What if every assumption fails?
- Either the idea is wrong or you mis-stated the audience. Try the same problem with a different segment before abandoning the space.
Sources & further reading
- Running Lean — Ash Maurya
- The Mom Test — Rob Fitzpatrick
- How to Talk to Users — Y Combinator Startup School