Plenty of plans that fail had someone in the room who saw the problem coming. They noticed it during planning. But they weighed up whether saying so was worth looking negative or slowing everyone down, and they kept quiet. So the plan got signed off with the flaw still inside it. Gary Klein is the psychologist who came up with the pre-mortem. He described the problem in his 2007 Harvard Business Review article on the project premortem: too many people “are reluctant to speak up about their reservations during the all-important planning phase.”
His fix is a short meeting with one unusual instruction. Before the plan is committed, the person running the meeting tells the team to imagine that it has already failed and asks each of them to write down why. That’s the whole method, and you could explain it in a minute. The hard part is getting people to write what they actually think, and then doing something with what they write.
So this guide spends most of its time on the hard part. You’ll see a complete premortem (Klein’s spelling) on a plan that has nothing to do with software. The plan is an office move, and you’ll see the reasons people gave, how the list was ranked and what changed in the plan as a result. After that we cover how to stop people softening the risks, what to do with the list once the meeting ends and when the exercise isn’t worth running. It’s a tool for before you commit, because once something has actually gone wrong you need a different one.
What is a pre-mortem, and how does imagining failure change what people say?
A pre-mortem is a short meeting held before a plan is committed, in which the team assumes the plan has already failed and writes down the reasons why. Klein’s own definition is a single line: “A premortem is the hypothetical opposite of a postmortem.” A post-mortem happens after something has gone wrong and asks what killed it. A pre-mortem asks the same question earlier, while the plan is still on paper and can still be changed.
That might sound like an ordinary risk discussion with a dramatic name, but the framing is doing real work. Klein draws the contrast himself: “Unlike a typical critiquing session, in which project 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.”
Here’s our explanation of why that small change matters. It’s our reasoning rather than a research finding. Even so, it’s the reason the meeting is built the way it is. If you ask a team “what could go wrong?” then anyone who answers is raising a worry about a plan their colleagues have worked hard on. It may even be a plan their boss chose. So they have to defend the worry, and they risk becoming the person who’s against the plan. But if you say “it failed, tell me why” then nobody is arguing against the plan any more. Everyone has the same job, which is to explain something that has supposedly already happened. That means the person with the most uncomfortable explanation is now simply the one doing the job well. Disagreement turns into a contribution.
There is some research behind the idea, and it’s worth stating carefully because it often gets stretched. Klein’s article cites a 1989 lab study by Deborah Mitchell, Jay Russo and Nancy Pennington. Jason Collins summarizes it in his notes on the premortem. People who imagined with certainty that an event had already happened came up with about 30% more reasons for it than people who treated it as something that merely might happen. But Collins adds that the researchers “did not assess the quality of the reasons.” So the finding is about how many reasons people produce, not whether those reasons were any good. That means Klein’s HBR wording claims more than the study measured, because he writes that the technique “increases the ability to correctly identify reasons for future outcomes by 30%.” We’ve read the paper’s abstract and Collins’ summary, but not the full paper. The abstract itself suggests that what changed people’s explanations was the certainty of the outcome, not the use of the past tense. That fits the way the meeting is run. The instruction doesn’t hedge: the plan has failed, full stop.
Klein’s format starts “after the team has been briefed on the plan,” during the planning phase. The best moment in practice is when the plan is nearly decided but nobody has signed anything that can’t be undone. Run it earlier and there’s no real plan to pick apart. Run it later and the answers can’t change anything. Picturing in detail how a plan will fail is a large part of strategic thinking, and a pre-mortem turns that into a habit of decision making a whole team can share. (For the wider skill set, see our guide to decision-making skills for managers.)
It’s also worth knowing what a pre-mortem isn’t. It isn’t a risk register, which is a running list you keep up for the life of a project; a pre-mortem is one meeting whose output may go into that register. It isn’t a retrospective either, because a retrospective looks back at work that’s finished.
A pre-mortem on an office move, run start to finish
This example is constructed for illustration. The company, the people and the figures are made up to show how a pre-mortem runs; they are not a real case or Risely data.
The company has 90 staff. About 55 work in customer service, including an early shift that starts at 7:30. About 20 work in finance and the rest are spread across IT, HR and facilities. The lease on its current office ends on 31 March and can’t be extended. So the company has already signed a lease on a smaller, cheaper office across town. The location is settled, but the layout isn’t yet.
The layout was designed by Duncan, the operations director. Because most staff now work in the office three days a week, he planned 60 desks for 90 people and a booking app to spread attendance across the week. His numbers came from the badge system: an average of 52 people a day over the past three months, with Monday the busiest day at 61. The fit-out budget is $250,000, which includes $20,000 held back for surprises. The move is planned for the weekend of 14 and 15 March, and the contract that fixes the desk count is due to be signed on 16 January.
The pre-mortem took place on 14 January, two days before that signature. That timing matters, because if it had happened after the contract was signed, nothing the team said about the layout could have been used.
Ten people were in the room. Duncan was there with Megan, the project manager running the move. Yvonne manages customer service, and Rafe is one of her team leads. Hugo is the finance controller and Shona works in accounts. Trevor is the IT manager, Lorna is the facilities coordinator and Fiona is from HR. The tenth was Kirsty, a customer service agent who’d been with the company for eight months and works the early shift. Yvonne suggested inviting her because nobody else at the table starts work before nine.
Setting it up
Because it was Duncan’s plan, he asked Fiona to run the session. He spent five minutes walking everyone through the layout, the dates and the budget before Fiona gave the instruction. She told the room it was now the end of April. The move had happened on 14 March, and it had gone so badly that the leadership team was talking about moving again. Everyone had eight minutes to write down every reason they could think of for why it had gone wrong, each on their own sheet of paper and without talking.
The silence is deliberate, and the section after this example explains why it matters so much. The short version is that writing alone first makes each list reflect what that person actually knows, rather than what the first confident voice in the room said.
The reading round
The reading round came next, and it followed Klein’s format. Megan read first because she was the project manager. Each person read one reason from their list that was different from anything already said. The round went on until every list was used up, and Duncan read last. Fiona wrote each reason on a flip chart, and there was a firm rule that nobody responded to a reason while the round was still going. That meant no “we’ve thought about that” and no “that won’t happen”. Twenty-six reasons went up in about twenty minutes.
Plenty of the twenty-six overlapped, and by the time Fiona had grouped the cards that described the same failure, ten distinct reasons were left. The clearest case was two cards that were really the same reason in different words. Lorna had written “not enough lockers”, while Rafe had written “agents have nowhere to keep their headsets”. They sound like separate complaints, but they come from the same fact: when nobody owns a desk, nothing can be left at a desk overnight. Customer service agents use their own headsets for hygiene reasons, so for them this was more than a nuisance. Fiona merged the two into a single reason about storage. That’s worth doing every time, because two cards about the same thing can make one problem look twice as important as it is.
Trevor’s reason got the biggest reaction in the room. He’d written that on Monday 16 March the phones didn’t work, and customers couldn’t get through. A customer service team with no phones on its first morning is a frightening picture. But when the list was ranked, this went near the bottom. That’s because the phone system runs in the cloud, so agents can take calls on their laptops from home if the office lines fail. The worst realistic version was a slow morning, and Trevor already had test calls booked for the Sunday afternoon. So the reason was dramatic but not likely, and the damage would be short.
Kirsty’s card was the opposite. It sounded like almost nothing when she read it: “Not sure the attendance numbers include everyone on earlies.” Once the round was over, Fiona went back to it and asked Kirsty to describe how that had played out, in the version of events where the move had failed. Kirsty explained what happens on the early shift. The first person to arrive props the side door open, and most of the people who come in after them walk straight through without swiping their badge. That’s about a dozen people on most mornings, and nobody else in the room arrives early enough to have seen it.
Megan was the first to say what that meant. The badge data was the basis for Duncan’s 60 desks, and the badge data was missing about a dozen people on most days. So the whole layout rested on a number that was wrong.
Ranking the list
Fiona asked Megan, Yvonne and Hugo to rank the list instead of Duncan. They asked two questions of each reason: how likely is this to happen, and how bad would it be if it did? Each reason got a rating of high, medium or low on both.
Duncan’s assumption about the desk count ended up at the top. That wasn’t because anyone thought he’d been careless. It was because Kirsty’s point made the failure close to certain on the busiest days. The damage would also arrive on the very first Monday, and the layout was about to be locked into a contract. It was the cheapest reason on the list to check, too. Duncan took it well at first and said, “Then let’s count.” He was less happy when it became clear what a bigger desk count would cost.
The table ranks the ten reasons left after grouping and shows the decision the team made about each one.
| Rank | Reason the move failed | Raised by | Likely | Damage | Decision and owner |
|---|---|---|---|---|---|
| 1 | Badge data misses most of the early shift, so 60 desks are too few on Mondays and Tuesdays | Kirsty | High | High | Fix: count attendance by hand for a week before the contract is signed (Lorna) |
| 2 | Fit-out finishes late and the company has nowhere to work on 1 April | Megan | Medium | High | Can’t fix: set a trigger date for a work-from-home fallback (Megan) |
| 3 | Finance can’t sit together during month-end close | Hugo | High | Medium | Fix: reserve a block of 8 desks for finance on close days (Hugo) |
| 4 | Staff on the east side quit over a commute that’s 25 minutes longer | Yvonne | Medium | Medium | Accept and write it down: count exits that mention the commute for six months (Fiona) |
| 5 | People book desks and don’t turn up, so the app shows the office full when it isn’t | Rafe | Medium | Medium | Watch: release unclaimed bookings at 10:00 (Lorna) |
| 6 | Nowhere to store belongings, especially personal headsets | Lorna and Rafe | High | Low | Fix: one locker per person (Lorna) |
| 7 | Customer service calls are too loud for finance to work beside them | Shona | Medium | Low | Fix: put finance at the far end of the floor (Lorna) |
| 8 | Laptops or screens are damaged in the move | Trevor | Low | Medium | Watch: the movers’ insurance covers it (Trevor) |
| 9 | Desk-booking app confuses people in the first week | Fiona | Medium | Low | Fix: two short demos before the move (Fiona) |
| 10 | Phones don’t work on the first Monday | Trevor | Low | Low | Watch: test calls on Sunday 15 March (Trevor) |
What changed in the plan
The hand count happened the next week, with Lorna and Kirsty counting heads at 7:45 and again at 10:30 each day. The real Monday peak was 71, and Tuesday’s was 66. Sixty desks would have been eleven short every Monday and six short every Tuesday, the two busiest days of the week.
That changed the plan itself. The team turned a planned project room into 10 more desks, which brought the total to 70. That cost $11,000 from the contingency. Yvonne also split customer service into two groups with different office days, so the whole department wasn’t trying to come in on a Monday. The two changes needed each other, because 70 desks on their own would still have been one short of the 71 people counted on Monday. Duncan argued for keeping the project room and splitting the teams more aggressively instead, and he lost that argument. He signed the revised contract on 23 January, a week later than planned.
The late fit-out was the reason nobody could fix. Megan had already asked the landlord in December whether the old lease could be extended, and the answer was no. So instead of a fix, the team set an early warning. If the contractor hadn’t confirmed completion by 6 March, Megan would switch customer service to working from home for up to two weeks. The switch was planned in detail in advance so it wouldn’t be improvised under pressure.
The commute was a risk the team chose to accept. The new office was picked because it’s cheaper, and that wasn’t going to change. But the team wrote the risk down along with the reason they’d accepted it. Fiona also agreed to count how many people who left in the next six months mentioned the journey. Writing it down was the point, because if people did leave over the commute, nobody could later say the problem had been missed.
After the move
The fit-out was signed off on 4 March and the move happened on schedule. Trevor’s phones worked on the Monday morning. The extra desks were used from the first week, and the Monday peak in April was 68.
Not everything went the way the list suggested. The lockers turned out to matter less than anyone expected, because most people took their laptops and headsets home every night anyway. By May about a third of the 90 lockers were empty. The session also missed a problem that the team’s own fix created. Converting the project room into desks left the office short of private space, so by the second week Yvonne’s team leads were holding their private catch-ups with agents in the kitchen. Nobody had listed that, partly because the decision to give up the room was made after the pre-mortem, not during it.
The commute is still open. Two people from the early shift had left by June, and one of them mentioned the journey. Fiona’s six-month count isn’t finished, so nobody yet knows whether accepting that risk was the right call.
How do you stop people softening the risks?
The reason that changed the plan most came from Kirsty, the most junior person in the room. It also arrived already softened: “Not sure the attendance numbers include everyone on earlies.” If Fiona had read it out and moved on, it would have sat on the flip chart as a vague worry about data quality. It would probably have been ranked somewhere in the middle. That’s a common pattern, and a lot of the skill in running a pre-mortem is getting reasons like that one out whole.
People soften risks because saying them costs something. A reservation about a plan can sound like a criticism of the person who made it, and that person is often more senior than you. Klein built the pre-mortem around this. His instruction is to write down every reason, “especially the kinds of things they ordinarily wouldn’t mention as potential problems, for fear of being impolitic.” The economist Richard Thaler put the same point more bluntly in a piece for Edge.org on the premortem: “The premortem procedure gives cover to a cowardly skeptic who otherwise might not speak up.” But the cover only works if the person running the session protects it, and these are the moves that do that.
Have everyone write before anyone speaks
Everyone writes on their own and in silence before any discussion starts. If people talk first, the first opinion stated becomes the anchor, and the most senior opinion tends to carry the most weight. Both of these are well-known cognitive biases in group decisions, and both make people quietly drop the reason they were going to give. A written list can’t be talked out of existence before it’s read. This is the same silent-brainstorm idea that works against groupthink in the workplace, applied to risks instead of ideas.
Keep the plan’s owner out of the ranking
Klein’s format has the project manager read first. That’s his design and it works, because it puts someone’s reasons on the table straight away and shows that the exercise is real. But the plan’s most senior owner is a different matter. Our editorial view is that the person with the most at stake in the plan should read last and should not be the one who ranks the list. That’s our opinion rather than a sourced finding. Klein warns about the “damn-the-torpedoes attitude often assumed by people who are overinvested in a project.” That’s a real risk here, because someone who designed the plan will often rank reasons about their own decisions lower without meaning to. It’s why Fiona ran the session and three other people did the ranking.
When the owner does speak, how they react matters as much as what they say. If the owner looks relieved when a reason is small and irritated when it’s about their own choices, everyone in the room notices. The next round is then likely to be softer. The best thing Duncan did in the example was say “then let’s count” when his own number was in question.
One reason per person, and no defending
Each person reads one new reason per round, and nobody replies while the round is going. That rule matters most for the plan’s owner. The moment someone answers a reason with “that’s already handled”, the exercise changes. It stops being a list of explanations and becomes a hearing in which every reason has to survive a defense. People notice that quickly and start leaving out anything they can’t prove. So questions and arguments wait until the ranking, when the whole list is visible.
Turn labels into stories
Some reasons arrive as labels like “timeline risk”, “stakeholder alignment” or “data might be off”. A label like that is usually a real concern wrapped in safe words. You can’t rank it, because you can’t tell how likely it is or what the damage would be. The fix is a follow-up question that asks for the story behind the label. You want the person to describe what happened in the version of events where the plan failed, rather than justify the worry. Fiona’s question to Kirsty was roughly “how did that play out, in the version where this sank us?” Asking for a story doesn’t put her on the spot the way asking her to prove it would.
Running it with a remote team
A remote or hybrid team can do the silent writing in a shared document or in a tool that hides names until everyone has finished. Anonymity makes it easier to write the uncomfortable reason. But it has a cost, because when nobody knows who wrote a vague card, nobody can ask the follow-up question. Kirsty’s card would have stayed vague if Fiona hadn’t been able to turn to her. A reasonable middle path is anonymous writing with an open invitation for authors to expand on their own cards during the ranking.
The people who soften risks are often the people who find it hardest to speak up in any meeting, and our guide to speaking up in meetings is written from their side. If you’re the one running the session, the hard part is staying quiet while your own plan gets picked apart. You can rehearse that with Merlin before the meeting: tell it what the plan is and which criticism you’d least like to hear. The meetings where people say what they think are usually run by managers who’ve spent months building psychological safety.
What do you do with the list after the meeting?
Klein says the project manager reviews the list after the session, “looking for ways to strengthen the plan.” That’s where a lot of pre-mortems quietly stop. The list gets typed up and emailed round, and then it’s forgotten. A few habits keep that from happening.
Every reason that survives the ranking needs one owner and a date. The hand count in the office move belonged to Lorna. It had to be finished by 23 January, which was the new signing date. A reason that belongs to “the team” doesn’t belong to anyone.
Each reason also needs an explicit decision: fix it, watch it or accept it. Fixing means changing the plan, as the team did with the desk count. Watching means setting an early-warning check, like Megan’s 6 March trigger date for the fit-out. Klein notes that the exercise “sensitizes the team to pick up early signs of trouble once the project gets under way,” and a named trigger turns that into something concrete. It tells people exactly what sign to look for and what happens when they see it.
Accepting a risk is a legitimate decision, but it has to be written down along with the reason. There are two reasons for that. If the risk does turn into a problem, the record shows that it was a deliberate choice. The record also gives you a way to check later whether the choice was right, which is what Fiona’s count of commute-related exits is for. A decision log is a good place for this. That post covers how to record decisions; this one is about producing the list that feeds them.
The last habit is to pick a checkpoint date to come back to the list and put it in the calendar on the day of the pre-mortem. The obvious dates for the office move were 6 March and the first Monday after the move.
If the plan fails anyway, working out why is a different job from imagining why. An 8D problem-solving report is built for that job.
When is a pre-mortem a waste of time?
None of the five top-ranking guides we read before writing this post covers this question, and the research doesn’t either. So what follows is our editorial judgment rather than sourced findings.
A pre-mortem is a waste of time when the plan is already committed and paid for, with nothing left that can change. At that point the exercise is theater, because people go through the motions and learn that their reasons go nowhere. If the office move contract had been signed before the session, Kirsty’s point about the badge data would have been interesting and useless.
It’s overkill when the decision is cheap or easy to undo. If you can try something next week and reverse it the week after, you’ll learn more by trying it than by imagining it failing. Our guide to reversible decisions covers how to tell the difference. When a decision is that easy to reverse, skip the meeting or do a five-minute version in your head.
It doesn’t work when nobody in the room knows the domain well enough to imagine how it fails. A pre-mortem draws on what people already know, so ten people guessing about something none of them has done will produce ten guesses. The list will also look more authoritative than it is.
It does harm when the leader won’t act on what comes out. This is worse than not asking at all, because it teaches the room that candor is collected and then ignored. The next time you ask, the cards are likely to be softer.
It gets unworkable when the group is too big. A group of twenty makes the reading round take so long that the later reasons get rushed. So split into smaller groups and merge the lists.
A pre-mortem also doesn’t tell you how likely anything is, because it produces reasons rather than probabilities. The ranking is still your judgment, and the example’s ranking could easily have been different with a different three people doing it.
Finally, the evidence itself is thin. Collins puts it plainly: “The academic literature on the premortem is thin.” Thaler clearly likes the technique, but he says the same in his Edge piece: “I know of no systematic study of their use. Organizations rarely allow such internal decision making to be observed and recorded.” A pre-mortem costs an hour, so it doesn’t need strong proof to be worth trying. But nobody can honestly promise you that it will catch the thing that sinks your plan.
If you have a plan due for sign-off this month, book an hour with the people closest to the work before you sign. Invite at least one person who sees the daily detail that the planners don’t. If the plan is yours, decide in advance who will rank the list.
Running the session well comes down to two moments: staying quiet while people explain how your plan failed, and replying to the hardest reason without defending yourself. Practice both with Merlin in Slack or Microsoft Teams before you walk into the room.