AI / Analysis & Decision Making
Split a Big Question Into Testable Pieces
Split down to pieces that answer yes/no or a number
My big question is: [question]. Break it into a set of independently testable sub-questions. Requirements: 1) 4–6 of them, each answerable with yes/no or a single number; 2) For each, what data you need, where to get it, and how long it takes; 3) Flag the ones where a “no” settles the big question; 4) Rank by information gained per unit of cost and name the first three to check; 5) Separate the two kinds of question I've blurred together — factual questions and choices — and say which is which. Do not answer the big question.
How to use
Answer your own list first. The sub-question you can't answer is the one actually blocking you.
Good for
- A big question has been sitting for two weeks, every meeting re-litigates the same ground, and nobody can say what to check next.
- You're asking for budget on a fuzzy direction; the reviewers will ask “on what basis,” and all you have is a general impression.
Not for
- The question is really a values choice — whether to leave this job, say — and splitting it into facts routes around the actual trade-off.
- You already know the data gap and have a specific experiment to run — decomposition becomes another meeting that produces nothing.
Example
Input
Big question: should we move our main acquisition channel from TikTok to Xiaohongshu? TikTok is 68% of new customers at ¥41 CAC; a ¥30k Xiaohongshu test came in at ¥63 CAC but with 1.8× the repeat rate.
What you get
The first testable sub-question is whether Xiaohongshu's CAC rises with scale — ¥30k is a small sample, ¥63 is probably cheap inventory, and if the answer is yes, 1.8× repeat doesn't cover the gap. The second is how much of TikTok's 68% comes from an already-saturated audience; that decides whether this is a choice or a forced move. Only after those two does “should we switch” become a real question — right now you'd be extrapolating one trial spend to the whole channel.
Common mistakes
- Stopping at the list: ten sub-questions written down, none investigated, and the next meeting reopens the big question.
- Splitting into sub-questions just as big as the original (“is there still room to grow on TikTok?”) — nothing is testable, so nothing was split.
Try tweaking it like this
- Add “mark each sub-question: would a wrong answer flip the decision, or only change the size?” — that cuts the ones that don't matter.
- Decompose the three questions you're most afraid of being asked instead — work backwards from the defense to what to check first.