How to run a pre-mortem: a 45-minute facilitation guide

By Meet Patel · 2026-10-03 · 7 min read

A pre-mortem is a meeting, held before a plan is committed, in which the team imagines the plan has already failed and each person writes down why. Gary Klein described the method in a 2007 Harvard Business Review article, and in a 2021 Psychology Today piece he puts the meeting at as little as 20 to 30 minutes. I would allow 45 for a plan with real money or a public date behind it. The output is a short list of causes, each with an owner, an early-warning signal and a change to the plan.

Take a six-person team (a hypothetical) about to launch a new pricing tier on 1 December. The plan reads well, the dates are agreed and nobody in the room objects. That silence means the doubts have not been spoken yet, and the meeting exists to hear them.

Why imagining failure works

Klein's article, “Performing a Project Premortem” (Harvard Business Review, September 2007), contrasts the method with an ordinary critique. In a typical session, team members are asked what might go wrong. The premortem “operates on the assumption that the ‘patient’ has died, and so asks what did go wrong.” The change of tense is the whole mechanism. Doubt becomes a contribution to a task, which is explaining a failure that has happened, and nobody has to criticize a colleague's plan to raise it.

Klein has also described the social problem the method addresses: “There is silent pressure in these meetings not to surface doubts.” After a plan has been presented and agreed, the cost of raising an objection feels high and the benefit feels low.

There is research behind the framing. Klein cites a 1989 study by Deborah Mitchell, Jay Russo and Nancy Pennington on what they called prospective hindsight. A summary of that work reports that imagining an event as already having occurred, rather than considering that it might, increased the number of reasons people generated for the outcome by about 30 percent. Two cautions apply. The finding measures how many reasons people produce, and it does not by itself show that pre-mortems improve project results. And the figure comes from a secondary summary, so I would quote it as roughly a third more reasons and no more precisely.

Before the meeting

Time it. Hold the pre-mortem when the plan is nearly final and the decision to commit is still open. Klein places it at a kickoff after the plan has been discussed. Once money is spent and announcements are made, a concern starts to sound like disloyalty.

Pick four to eight people. Include the people who will build it, the people who will sell or support it, one habitual skeptic and one person who will be affected by the result but has not been in the planning. Their views of what could break differ, and that variety produces the list.

Prepare one page. The goal, the date, the three assumptions the plan depends on most, and who does what. A longer document turns the meeting into a reading exercise.

Choose a facilitator who does not own the plan. An owner who runs the session will, without meaning to, defend the plan while collecting objections to it. If the owner must facilitate, they take part last.

The 45-minute facilitation

1. Present the plan (5 minutes). Walk through the one page. Take clarifying questions only. Do not defend or debate.
2. Announce the failure (3 minutes). Say it in the past tense and with certainty: “It is three months after launch. The plan has failed badly. In the next few minutes, write down why.” Klein frames this as looking into a crystal ball that shows the failure, so the team is explaining something that has happened.
3. Write in silence (5 minutes). Everyone writes reasons on their own, one per note. Klein suggests two minutes. I would allow five when the stakes are high, and ask for at least five reasons each. Silence matters because the first person to speak otherwise anchors everyone else.
4. Read the reasons around the room (12 minutes). Each person reads their single most important reason, and the facilitator posts it on a shared board. Go round again until the notes run out. The leader or plan owner goes first, which models candor for everyone after them. Allow no rebuttals.
5. Cluster and vote (10 minutes). Group duplicates. Give each person three votes to place on the causes they judge most likely or most damaging. Keep the top three to five.
6. Turn causes into changes (10 minutes). For each top cause, write four things: a change to the plan, an early-warning signal that would show it happening, a named owner and a date.

At the end you have what Klein calls a list of showstoppers, and a team that has been prepared to look for them.

Prompts that draw out better reasons

Some teams freeze at the blank note. Put three or four prompts on the board before the silent writing starts. These four are general enough to start with.

• What did we assume about customers that turned out to be wrong?
• Which dependency, on a person, a vendor or a system, let us down?
• What did we leave out of the plan because it was awkward to discuss?
• What changed outside the company that we did not allow for?

The third prompt is likely to be the most productive, because it invites the topics the planning meeting avoided.

A worked example

Here is the pricing launch, with illustrative contents. After the silent writing, the team's notes cluster into five causes:

• Existing customers move down to the cheaper tier and revenue falls.
• The billing system cannot prorate mid-cycle upgrades correctly.
• Support is not trained, and ticket volume spikes in launch week.
• Sales keeps quoting the old price for two weeks.
• The usage limit on the middle tier is hit by many customers in the first week.

The vote puts downgrades first, proration second and support readiness third. The team then writes its actions.

• Downgrades. Change: offer existing customers a six-month price lock instead of an automatic move. Signal: more than 10 percent of existing customers asking to move to the lower tier by the end of week two (an illustrative threshold). Owner: head of growth. Date: 20 November.
• Proration. Change: test five mid-cycle upgrades in the billing sandbox before launch. Signal: any invoice that differs from the hand calculation. Owner: finance lead. Date: 24 November.
• Support readiness. Change: a 30-minute briefing and a five-question reference sheet. Signal: tickets per 100 customers on day two. Owner: support lead. Date: 28 November.

Each signal gives the team something to watch after launch, and that links the pre-mortem to the problem of metrics that drift before anyone notices. Record the output as well. A decision log entry that names the top causes and the changes made lets the team compare them with what happened.

Variations for remote teams and solo work

A remote team can use a shared document with a timer. Each person types reasons into their own section, collapsed until time is up, and the facilitator then reads them round the room in the same order as an in-person session. The silence is the part to protect, so avoid a live chat thread where early replies steer the rest.

A solo operator can run the same sequence alone on paper the evening before a commitment. A single mind rarely holds every failure path, so hand the list to one trusted outsider and ask them to add what you missed. Repeat the exercise at each major milestone, such as a budget release or a hiring plan, because the failure modes change as the plan grows.

Mistakes that waste the meeting

• Holding it after the commitment. By then the choices that matter have been made, and the list becomes a record of regret.
• Letting the plan's owner defend it. One rebuttal teaches the room that objections have a price, and the flow of reasons stops.
• Skipping the silent writing. Without it, the first confident voice sets the agenda and the quieter people confirm it.
• Naming people. A cause such as “the dependency on the payments vendor has no backup” gives the team something to fix. A cause that names a colleague ends the candor and fixes nothing.
• Ending with a list. A list with no owners and dates is a record of worry. The last ten minutes are the part that changes the plan.
• Treating votes as probabilities. Dots show which concerns the room holds most strongly, which is a different thing from how likely each one is. If one cause carries a large downside, handle it even when it received few votes. The asymmetric bet framework is a useful lens for that call.

What a pre-mortem can and cannot do

It surfaces concerns that people already hold and gives them a safe route into the plan. It cannot produce knowledge that nobody in the room has, and it does not replace testing the assumptions it uncovers. Treat the output as a prompt for the next round of checks.

A plan that survived an imagined failure can still fail. The difference is that some of the ways it might fail now have a name, an owner and a signal someone is watching.

Written by Meet Patel — startup operator and growth strategist in Dubai.

Read on themeetpatel.com