Risky Assumption De-Risking
- Difficulty
- Moderate
- Time to result
- ~weeks to results
- Steps
- 6
- Confidence
- —
A method for testing slow, expensive ideas in a week or two. Most of the ideas that excite founders are hard: publish a book, build a product, run a big campaign. Rather than committing years to find out, you break the idea into its moving parts, name every risky assumption inside it, and isolate the one that would kill the idea if it were false. Then you find the cheapest possible proxy test for that single assumption — often just emailing people who have already done the thing and asking what actually happened. In a week you've either de-risked the assumption or ruled out a bad idea. The final step is writing down what you learned so the organisation doesn't have to relearn it.
Origin
Extracted from Deep Dive with Ali Abdaal
How to run it
- 1
Write out the idea and its moving parts
State the idea in full, including why you think it will work commercially. Example: publish a book about productivity, because readers will then sign up to the course. Lay out every moving part that has to hold for the idea to pay off.
- 2
List every risky assumption
For the book example: that you can write a book reasonably quickly, that it won't distract you from running the business, that people will want to buy it, that they'll like it, and that people who like it will then sign up for the course. Most big ideas hide five or more assumptions like this.
- 3
Isolate the assumption that decides the idea
You can't test all of them without building the thing. Pick the one whose failure would make the whole idea worthless — here, whether book readers actually convert into course buyers.
Watch out If you try to test everything at once, the only valid test is doing the project — which is exactly what you were trying to avoid.
- 4
Design a test you can run in a week or two
Ask what would tell you the answer fastest: how hard writing is, whether people respond to a particular title or topic, whether readers go on to buy a course. The test needs to be days, not quarters.
- 5
Use people who already ran the experiment
Ask who you know who has both a book and a course, then email them and ask directly to what extent the book contributed to course sales. Their answers, and the reasons behind the split, are usually more informative than a synthetic test.
Pro tip Look for the variable that explains disagreement between respondents — in this case, whether they already had a YouTube channel.
Watch out A handful of data points is directional, not proof; look for the pattern that explains the difference.
- 6
Document the result and the learning
Record the experiment and what you learned. Otherwise the same idea gets re-proposed and the same mistake made again, by you or by someone else in the team.
Watch out If people fear looking bad, they won't discuss their mistakes and the learning never propagates.
In the wild
Ali's example: he sells a productivity community and course, so publishing a productivity book looks like a great accretive idea — except it takes three and a half years. The risky assumptions are that he can write it quickly, that it won't distract him from the business, that people will buy it, like it, and then take the course. Instead of writing the book, he isolates the last assumption and emails people he knows who have both a book and a course. Some said the book did basically nothing because they already had a YouTube channel; others said it was massive precisely because they didn't.
→ A three-and-a-half-year bet gets a directional answer in a week, with the deciding variable identified.
Common mistakes
Confusing wanting to do it with a business case
Sometimes you just want to write the book, which is fine — but don't dress that up as a growth lever and skip the test. Separate the personal motivation from the commercial assumption and test the commercial one honestly.
Skipping the write-up of what you learned
Teams often only remember an experiment two years later, after repeating the mistake. Documenting experiments and learnings is an explicit part of the process, not optional overhead.
A culture where mistakes can't be discussed
Someone afraid of a bad grade or getting fired won't talk about their mistake, so nobody learns from it and someone else goes and makes it again.
From the transcript
“so the way you'll do that is you'll look at this idea and you'll say okay all the thing the moving parts of this what…”
“so for each of your experiments that you're most excited about you figure out what's the quickest way we can test this”
“actually that was part of the experiment process is document your experiments done and document your learnings”
From the episode
How I’m Scaling My Business from 7 to 8 Figures