Bottom line
A pre-mortem is a short meeting, usually 60 to 90 minutes, in which your team assumes a decision has already failed and writes down why, before anyone signs off. It works because imagining failure as a fact makes risks easier to name: research cited by Gary Klein in Harvard Business Review (2007) found that this "prospective hindsight" increases the ability to correctly identify reasons for future outcomes by 30%. Use it on the few decisions that are expensive to reverse, after the options are clear and before final approval.
A normal risk review misses the risks people are reluctant to say out loud.
Once a plan has a sponsor and a slide deck, raising doubts feels like disloyalty, and the risk list fills up with safe items such as "the economy" or "hiring is hard". The pre-mortem changes the question from "what could go wrong?" to "it went wrong, explain why", which gives everyone permission to be specific. Klein, who developed the method, and his colleagues tested it: in a 2010 study, 178 college students critiqued a plan for keeping students safe during a hypothetical H1N1 flu outbreak, and those in the pre-mortem condition showed the greatest reduction in overconfidence compared with simple critiques or pros-and-cons lists (Klein, Psychology Today, 2021). Both findings come from lab studies, so treat them as evidence that the framing surfaces more and better doubts, not as proof that projects succeed 30% more often.
Run a pre-mortem only where a wrong call would be hard to undo.
Good candidates are market entry, a senior hire, an acquisition or a succession step, a price change across your whole customer base, and large build-vs-buy commitments. A cheap, reversible test does not need one. Timing matters as much as selection: run it once you have a preferred option, so there is something concrete to attack, but before sign-off, while changing the plan is still cheap. If you run it after the decision is announced, people defend the plan instead of testing it.
Use a six-step framework and keep it to one session.
- Frame the failure in one sentence with a date: "It is November 2027. We entered France a year ago and it was a clear failure."
- Write silently for ten minutes. Each person lists every reason they can think of, independently. Invite at least one person who did not build the plan.
- Share in rounds, one reason per person per round, with the most senior person speaking last. No debate yet.
- Cluster the reasons and score each cluster on likelihood, impact and how early you would notice it.
- Act on the top three to five: change the plan, add a safeguard, or set a stop rule with a named signal and check date.
- Record what changed, who owns each safeguard and when you will check, in the same place as the decision itself.
More on Decision Science — every week.
Join 2,400+ leaders getting the decision intelligence briefing.
The method does not have to slow you down.
A common objection from founders is that there is no time for another meeting. A 2017 Michigan Technological University thesis tested the pre-mortem with 52 Army cadets working in teams on six time-constrained field exercises; when teams used it, they had fewer fouls and less fixation, with no change in planning and execution time. It is a small, military sample, but the practical point carries over: the cost is one focused hour, and the payoff is a plan that has been attacked by its own authors. That matters because most organizations are not confident in their decision process to begin with. In a 2018 McKinsey survey of 1,228 managers, only 20% said their organizations excel at decision making (a self-reported figure).
Hypothetical example: a market-entry decision.
Imagine a 90-person B2B software company in Munich planning to open a sales office in France with two senior hires and a local reseller. In the pre-mortem, the team's failure stories cluster around four themes: the reseller competes with the direct sales team for the same accounts; the two hires take six months to land, so the first year is lost; French public-sector buyers expect local-language support and contract terms the product team has not scoped; and the board member sponsoring the move steps back. The team keeps the goal but changes the plan: one hire plus a two-quarter reseller pilot with clear account rules, a support scope agreed before launch, and a stop rule if fewer than an agreed number of qualified opportunities exist by the end of the second quarter. This is an illustration, not a customer case.
Avoid the mistakes that turn a pre-mortem into theatre.
Do not let it become a blame session about past projects. Do not accept generic risks that apply to every plan; ask "what exactly happened, and when?" Do not run it with only the plan's authors in the room. Above all, do not stop at the list: a pre-mortem that changes nothing in the plan was a brainstorm, not a decision tool. And it does not replace legal, tax or financial due diligence; this article is general guidance, not professional advice.
AI can widen the failure stories, but the team still owns the judgment.
Small leadership teams often share the same blind spots, and AI is useful for generating failure stories from perspectives nobody in the room holds, such as a skeptical CFO, a key customer or a competitor. If you want that challenge in a structured form, NeuroAgent's council includes a dedicated Challenge stage and an independent auditor that checks the result against your constraints; you can try it free or bring a high-stakes call to a Decision Sprint. Either way, the goal is not to predict the future. It is to make sure the failure you fear most is one you have already planned for.
Research context
In a McKinsey survey, 72% of senior executives said bad strategic decisions were as frequent as good ones or were the prevailing norm.
McKinsey & Company: Untangling your organization's decision making (2017)This independently published statistic provides context for the topic; it does not substantiate every claim in this essay.
One decision framework, every week.
The week's best articles on making and executing high-stakes decisions, backed by sourced research.