Ask a team to list the risks in a plan and you get the three that are already on the register. Tell the same team the plan has failed and ask why, and the list gets longer and more specific.
Code: Pre-Mortem-Skill, GPL-3.0 licensed.
The effect has a name. Prospective hindsight is the mental move of treating an uncertain outcome as a fact that already happened, then explaining it. Research on the technique reports roughly a thirty percent increase in the number of accurate causes identified, compared with asking what might go wrong.
That number is why the technique survives in decision practice despite sounding like a trick of wording.
What changes in the head
A risk question asks the reader to predict. Prediction invites hedging, because being wrong about the future is embarrassing.
An explanation asks the reader to reconstruct. There is no future to be wrong about; the event is given. The mind switches from estimating probability to building a story, and stories come with details: who dropped the ball, which dependency slipped, what the customer actually said in week three.
That switch is also the danger. A confident story about a failure that never happens feels as real as one about a failure that does, and the technique generates both without distinguishing them.
Why the count goes up
Three mechanisms explain most of the gain.
The first is permission. Pessimism about an uncertain plan reads as disloyalty. Pessimism about a past event reads as analysis, and the same person says the same thing in a different voice.
The second is specificity. “The integration might be hard” is a prediction. “The integration failed because the vendor’s sandbox reset every night and nobody documented it” is a cause, and it names a fix.
The third is memory. Asked to explain a failure, people reach for similar failures they have seen. That pulls in base rates: the migration that overran, the hire that left in month four, the pricing change that lost the mid-tier. Those are the boring causes that kill more plans than any dramatic risk.
Building it into a tool
The Pre-Mortem skill packages the protocol as a prompt file with seven steps: input check, failure frame, failure generation with likelihood and impact scores, parallel deep dives, synthesis, a commitment ask, and artifacts.
Two design choices in that list matter for anyone building a similar tool.
Boring failures are mandatory. The protocol requires at least two mundane causes, because a model asked for failure reasons produces dramatic ones by default: market collapse, competitor pivot, regulatory change. Those are real, and they are not what usually happens.
A gap check closes the report. The synthesis has to name what the analysis probably missed. A premortem that claims completeness is worse than one that admits its blind spots, since the reader stops looking.
The commitment step is the other unusual part. The output ends with one concrete action and a day, and a refusal is recorded along with the failure mode that was dismissed. Recording refusals turns the exercise into a small archive of decisions that can be revisited.
Where a model helps and where it does not
A model is good at generating twenty specific failure reasons from a described plan, and at writing the early warning sign for each one. It is fast, tireless, and does not flinch at saying the plan is weak.
It is bad at knowing the organisation. Whether a team will actually lose interest in month three depends on people the model has never met, and a confident paragraph about team morale is a guess with a nice font.
The honest division: the model produces the candidate list and the structure, and the people in the room decide which candidates are real. Anything else turns a useful exercise into a well-formatted fiction.
The habit underneath
Treating a decision as already failed is one instance of a broader habit: run the check before the launch, not after. Audit the head of a page before it ships with a head audit, count the URLs in a map with a sitemap check, and write down what you expect the numbers to be.
Background: Gary Klein, Performing a Project Premortem, and Daniel Kahneman’s work on the outside view.