Guides
Decision frameworks

The pre-mortem template, and what it is for

A pre-mortem asks the team to imagine the project has already failed. Then you work backward to what went wrong. The standard version stops there and changes nothing. Annie Duke treats the output as raw material. Every failure signal becomes a kill criterion with an action attached. You decide it before anyone has skin in the outcome.

Why this matters. Teams run pre-mortems and feel sober about it. Then they build what they were always going to build. The exercise produces insight that never reaches a decision.

Ask Gavel, free

Bring the project you are about to commit to. Gavel runs the pre-mortem with you and turns the signals into criteria you can act on.

30%

more reasons for potential failure surface when you pair a pre-mortem with backcasting, per Gabriele Oettingen's research on mental contrasting. Imagining only success produces blind spots; imagining only failure produces pessimism with no roadmap.

Gabriele Oettingen's research, cited by Annie Duke Annie Duke, Thinking in Bets

The short answer

The pre-mortem template

Send this as a written prompt before anyone meets. Independent answers first, discussion second.

  1. 1

    The prompt

    It is [date, six to twelve months out]. We committed to [project] and it failed. Looking back, there were early signals. What were they?

  2. 2

    Answer alone, in writing

    Every participant answers alone before the group talks. Discussion first breeds conformity, and the most useful signal stays unsaid.

  3. 3

    Also run it in reverse

    Ask the same question about succeeding. Backcasting plus pre-mortem gives you both the roadmap and the risks; either alone is half a picture.

  4. 4

    Separate what you control

    Sort every factor into within your control and outside it. Only the first list is actionable; the second is what you monitor.

  5. 5

    Convert each signal into a criterion

    Turn the signal into a measurable condition with a date. Then attach an action: kill, pivot or test.

The cited playbook

How to run it so it changes something

The difference between a pre-mortem that works and one that does not is in what happens after the list.

  1. 1

    Treat the pre-mortem as a signal generator, not a plan review

    Annie Duke's observation is that running a pre-mortem does not change the plan much. Its real value is the list of things you would observe on the way down, because each of those can become a pre-committed exit condition. Run it to find signals, not to feel thorough.

    Annie Duke · Annie Duke on Lenny's Podcast
  2. 2

    Collect answers independently before anyone speaks

    Duke uses nominal group technique for this. When discovery and discussion happen in the same room at the same time, the first confident voice anchors everyone else. The quietest useful signal is never raised. Each person answers alone in writing first. Then discuss the combined list.

    Annie Duke · Annie Duke, Lenny's Podcast interview
  3. 3

    Pair it with a backcast so you get a roadmap too

    Backcasting imagines the same future date with the project succeeding, and works backward to what made it happen. Duke's framing is that the two are complements: a pre-mortem alone breeds pessimism without direction, a backcast alone breeds false confidence. Run both, then sort every factor by whether you control it.

    Annie Duke · Annie Duke, Thinking in Bets
  4. 4

    Convert every signal into a kill criterion with an action

    A signal with no attached response is a discussion. Duke's worked example with a sales team turned three pre-mortem signals into standing rules. One rule: kill any deal where the buyer refuses a demo and only wants to talk price. Give each criterion a metric, a date and a committed action.

    Annie Duke · Annie Duke on Lenny's Podcast

Where experts disagree

Where operators disagree: what is the exercise for?

Annie Duke

says the pre-mortem does not improve the plan much, and that expecting it to is why the exercise so often changes nothing. Its value is generating the signals you will pre-commit to acting on. That is a decision-hygiene move, not a planning move.

Shane Parrish

treats scenario exploration as improving the decision itself. Imagining at least three futures and asking then what again and again surfaces second-order consequences you would otherwise miss. That changes which option you pick, not just when you would abandon it.

Both are right about different stages. Parrish's version belongs before you choose, when options are still open. Duke's belongs once you have made the choice, and the remaining question is what would make you reverse it. Running only Parrish leaves you with no exit condition; running only Duke means you never questioned the plan.

What founders say

What founders actually say

“Stop second-guessing”
Founder JTBD, Gavel customer research
“Move from paralysis to action”
Founder JTBD, Gavel customer research

Verbatim user quotes from public forums, sourced, not paraphrased.

FAQ

Pre-mortem questions, answered

What is a pre-mortem?

A pre-mortem asks a team to imagine a project has already failed at some future date, then work backward to the reasons. You run it before committing, unlike a post-mortem. Its purpose is to surface risks while they are still cheap to act on.

What is the pre-mortem template?

One prompt, answered alone in writing: it is [date] and this project failed. Looking back, there were early signals. What were they? Then sort the answers by what you control, and convert each signal into a measurable condition with an action attached.

How is a pre-mortem different from a risk assessment?

A risk register lists what might go wrong and assigns a probability. A pre-mortem starts from the assumption that it already went wrong. That surfaces more specific failure modes, because the mind finds it easier to explain a fact than to forecast one.

Does a pre-mortem actually change anything?

On its own, not much. Annie Duke's finding is that teams run one and then proceed with the original plan. It changes something when you convert each surfaced signal into a kill criterion with a committed action. Decide that before anyone has a stake in the outcome.

Bring your actual numbers. Get a cited answer you can defend.

Try Gavel free

Related guides

Keep exploring