Assumption Collision Test
Test the riskiest premise with a real person before building the business
- Difficulty
- Starter
- Time to result
- ~days to results
- Steps
- 5
- Confidence
- 99%
Marc Randolph starts from the belief that every idea is bad in some unknown way. Instead of defending it with a business plan, identify the assumption most likely to kill it and create a quick, cheap encounter with reality. The test need not be scalable or even a viable product; it only needs to produce relevant behaviour. Mailing a compact disc tested whether a DVD-sized object could travel safely and cheaply, while a handwritten dorm-room sign could test whether anyone wanted to borrow clothes before a marketplace was coded. If the premise fails, stop or change direction with little sunk cost. If people respond, use the next observed friction to design the next test. The output is evidence about what to test or build next, not proof that the whole venture will succeed.
Origin
Randolph and Reed Hastings turned their car around during a commute, mailed a used music CD to Hastings's home, and saw it arrive intact the next morning for one stamp. That crude test kept the DVD-by-mail idea alive without a business plan or product build.
Core principles
- 01Assume every new idea contains a fatal flaw
- 02Real behaviour is stronger evidence than encouragement
- 03Test the riskiest premise before building supporting machinery
- 04A cheap rejection preserves time for the next idea
How to run it
- 1
Assume a hidden flaw
Treat enthusiasm as a hypothesis, not validation. Write down why the idea might fail before imagining its full potential.
Pro tip Start with the assumption whose failure would make the rest irrelevant.
Watch out Praise from a partner or friend is not customer behaviour.
- 2
Define the first behaviour
Choose the smallest real action a potential user must take for the premise to remain plausible. Make the signal observable and specific.
Watch out Do not use vague interest as a substitute for action.
- 3
Build an unviable test
Create the quickest manual or improvised version that can elicit that behaviour. Ignore scale and polish at this stage.
Pro tip Use ordinary materials or manual fulfilment when they answer the question.
Watch out A test should still represent the premise fairly enough to produce useful evidence.
- 4
Collide with reality
Expose the test to real potential users and observe what happens. Capture refusals, confusion and unexpected uses as well as positive responses.
Watch out Do not coach participants toward the answer you want.
- 5
Choose the next move
Abandon a disproven premise, revise around the friction, or run a deeper test when the signal is promising. Keep each decision tied to observed evidence.
Pro tip Let each test expose the next uncertainty.
Watch out One successful premise test does not validate the whole business.
In the wild
Randolph and Hastings could not find a DVD, so they bought a used music CD of the same size and weight, placed it in a greeting-card envelope, and mailed it to Hastings. It arrived intact in less than a day for the cost of a stamp.
→ The delivery premise looked feasible enough to justify exploring DVD rental by mail.
Randolph advised a student considering a peer-to-peer clothing marketplace to tape a sign outside her dorm asking whether anyone wanted to borrow her clothes. Knocks, choices, fit problems, damage and cleaning would reveal demand and operating friction without code or fundraising.
→ The student could learn whether the first premise and later practical assumptions held before dropping out or hiring a team.
Common mistakes
Building the imagined empire
A detailed business plan can make an untested idea feel real while leaving its most basic customer assumption untouched.
Testing the whole business at once
Large tests obscure which assumption produced the result and increase the cost of learning.
Is it for you?
Best for
It is best for founders evaluating an early idea whose first uncertainty can be tested manually.
Not ideal for
It is not ideal when a crude test would expose people to material safety, legal or financial risk.
From the transcript
“every idea is bad we just don't know why they're bad yet”
“how can I quickly cheaply and easily Collide this idea with a real person”
“we did not go to the office and do a business plan”
From the episode
Former Netflix CEO: “They're Lying To You About Hard Work!” Building a $278 Billion Company Wasn’t Built On Hard Work!