AI / Analysis & Decision Making
Pre-Mortem
Assume it has already failed, then work backwards
Assume [project/plan] has failed completely six months from now. Write the post-mortem: 1) The 5 most likely reasons for failure, ordered by probability; 2) For each reason, what the early signal looks like today; 3) Which causes can be prevented right now; 4) If you could do only one thing to reduce the chance of failure, what would it be.
How to use
Name a failure date and define what 'failed' means in one line; otherwise the causes stay generic.
Good for
- The project is approved, the budget is signed, the schedule is out — you can still change direction, but nobody on the team is willing to pour cold water anymore.
- The last review before launch: everyone is discussing how to make it better, and no one is asking what total failure would look like.
Not for
- Nothing has taken shape yet and it's still an idea — the “reasons it failed in six months” come out generic (“the need wasn't real”) because you haven't made any testable choices yet.
- You just went through a real failure and are doing an incident review — work backwards from evidence here; hypothetical causes will cover up the actual one.
Example
Input
We plan to launch a paid membership (CNY 39/month) in six months. We have 8,000 free users and 2,600 monthly actives.
What you get
You want five causes ranked by probability, each tied to a signal you can look at today — for example, “30-day retention is 11%, so the habit isn't there and payment has nothing to attach to.” It should end with one preventive action you can do this week at the lowest cost, not a risk register to file away.
Common mistakes
- Treating the list as a risk register: you save it and never attach a check date to any item, and six months later you find you've just been ticking boxes on your own list.
- Asking only “why did it fail” and never “what does failing look like first” — you get correct but unobservable causes (“the market is competitive”) that you can't act on.
Try tweaking it like this
- Change “failed completely in six months” to “still alive in six months but miserable” — most projects don't die suddenly, they bleed, and this framing digs up truer causes.
- Add “which of these failure causes will we have created ourselves in the first two weeks” — attribution moves from the outside world to actions you control.