Judgment Playbook

Pre-mortem

Assume the decision already failed and work backward while prevention is still possible.

5 minute readtaught in course chapter 9
UncertaintyCommitment

What it is

A pre-mortem is a short exercise run before a plan is approved. The team is told the plan has already failed, and each member writes down why, privately, before anybody speaks. What comes out is a list of specific causes of failure, ranked by plausibility, plus at least one change to the plan or to what the team agrees to watch.

Where it comes from

The method is Gary Klein's. Klein is a cognitive psychologist who studies how experts decide under time pressure, and he published it as "Performing a Project Premortem" in Harvard Business Review, September 2007. Writing in Psychology Today in January 2021, he said he had developed the method three decades earlier to cut down on problems in his own company's projects.

The idea underneath it is older and has a separate source. Klein's article credits prospective hindsight, the trick of imagining an event as already past, to a 1989 paper by Deborah J. Mitchell of the Wharton School, J. Edward Russo of Cornell and Nancy Pennington of the University of Colorado, published in the Journal of Behavioral Decision Making. Klein reports their result as a 30 per cent gain in the ability to identify reasons for future outcomes. The paper is narrower than that number suggests: its authors concluded that certainty about the outcome, more than the shift in time, changed how people explained events. The exercise itself was later tested by Beth Veinott, Klein and Sterling Wiggins, with 178 participants critiquing a pandemic response plan, and the pre-mortem group's confidence in the plan fell furthest.

What it corrects

A plan has been presented, the sponsor is visibly committed, and the meeting is nearly over. Someone in the room holds a concrete doubt: a dependency they know is fragile, or a deadline that collides with another cycle. They say nothing. The prompt they were given, "any concerns?", asks them to evaluate a plan rather than to produce a failure, so it returns adjectives instead of causes. Speaking also costs them something. The doubt sounds like criticism of a colleague's work, and it will look foolish in public if the project goes well. Ordinary care fixes neither problem. Care makes people study the plan in front of them more closely, and that plan contains the assumptions they should be questioning. Klein's move is to change the question. Failure is announced as a fact, so the task stops being judgment of a plan and becomes explanation of an outcome, which is work an expert can do quickly and in detail.

How it works

  1. Brief the plan in full, so that everyone argues about the same plan.
  2. Announce the failure as a completed fact at a stated date: the launch slipped nine months and the product was withdrawn.
  3. Give two or three minutes of silent, independent writing. Everyone lists causes, including the ones they would normally leave unsaid.
  4. Go round the table, one cause per person, the leader first, and record every cause before anybody debates one.
  5. Merge the duplicates, then rank by plausibility rather than by drama.
  6. Convert each of the top causes into a change: a prevention, a tripwire with a named signal and an owner, or a contingency you are willing to fund.

Worked example

Klein's 2007 article records a session on a project to put advanced planning algorithms into the hands of military air-campaign planners. One team member had sat silent through a long kickoff meeting. In the pre-mortem he supplied a cause: an algorithm would not fit comfortably on the laptops used in the field, so the software would run for hours when planners needed an answer in minutes, and the project would be impractical. The developers then admitted they had already built a shortcut and had been reluctant to mention it. The shortcut was substituted, and the project succeeded. Nothing was discovered in that room. Two pieces of knowledge already inside the team were put on the table, by an exercise that Klein says takes twenty to thirty minutes.

In a Business Case Weekly case

The Rolls-Royce case closes on a financing decision in mid-2020: revenue down by roughly a third, a loss heading for £4 billion, 9,000 job cuts already announced, and a choice among a deeply discounted rights issue, new debt, state support and selling divisions. A pre-mortem there is not pointed at the pandemic, which has already happened. It is pointed at the mix you choose. Fix a date eighteen months out, state that the company had to return to the market for money a second time on worse terms, and name the part of your own mix that produced that. The causes worth writing down are the ones only this company could generate: how it is paid, what its cash promise to airlines commits it to, and who is left holding the risk in each option.

In your answer

  • "Assume it is [date] and this decision has failed. The most likely cause is ..."
  • "That failure runs through a single assumption: ..."
  • "The earliest signal of it would be ..., and I would expect to see it by ..."
  • "So I would change the plan now by ..., and keep ... in reserve."

Common misuse

The usual counterfeit is the disaster list: a page of large, generic catastrophes, recorded and filed. It carries the method's name and does none of its work, because a cause that would fit any project cannot improve this one. Two tests tell them apart. Could the cause you wrote stand unchanged in a competitor's plan? Then it is not yet specific. Did any line on the list change the plan, the budget or the monitoring? If not, the meeting produced anxiety rather than judgment.

References

Red team

Next method

Give a protected opponent the job of breaking the plan rather than polishing it.

Open next tool