We were brought into a company late last year to review what had gone wrong on a launch that, by every executive's account, had been "thoroughly de-risked." The phrase came up four times in our intake call. Among the de-risking exercises was a premortem, dutifully scheduled, well-attended, and — we eventually established — almost entirely useless. We have the file. The risks identified were "competitive dynamics," "regulatory environment," "macroeconomic conditions," and "execution challenges." Every one of them was true. None of them was specific enough to act on. None of them was the thing that actually killed the launch, which was that the integration committee met monthly while the team needed weekly decisions, and the year was burned arguing about decision rights.
The team had not skipped the premortem. They had done it wrong. This is, in our experience, the more common failure mode by a wide margin. We have written elsewhere about why the premortem is free, and why most teams skip it. This essay is for the teams that don't skip it — and still get nothing from it. The technique is good. The facilitation, in most rooms, is bad. Five specific failure modes do most of the damage.
Failure mode one: opening with the senior person
The premortem, as Gary Klein originally framed it, depends on a small but irreducible piece of mechanics. The room is asked, individually, to imagine the project has failed two years from now, and to write down why. Individually. In writing. Before anyone speaks.
Most facilitators do not do this. They open the meeting, frame the exercise, and then — because they were taught to begin meetings this way — turn to the SVP and ask, "what do you see as the biggest risk?" The SVP, who is senior for a reason, offers an articulate answer. The room hears it. Every subsequent contribution is now, consciously or not, calibrated against that answer. The premortem has become a conversation about the SVP's risk, with footnotes.
This is not an obscure point about anchoring effects. It is the foundational mechanic of the technique. The premortem works because it pulls into the open the things that individuals in the room can see and that the group has been suppressing. The moment you let the group dynamic in before the individual writing, the suppression returns and the technique stops functioning. Most premortem failures we audit trace, in part, to this opening move.
The fix is mechanical. The facilitator hands out paper. The room writes. Nobody speaks. The senior person writes too — and writes first, in fact, because the senior person is the one whose unwillingness to write would otherwise license everyone else's. Only after the writing is done does the conversation begin.
Failure mode two: asking the room out loud instead of forcing private writing
Related, but distinct. Even facilitators who avoid opening with the senior person often skip the writing step entirely. They go around the table. "Bob, what could go wrong? Linda, what about you?" The room produces a list, the list goes on the whiteboard, everyone feels they have done the exercise.
They have not. They have done a risk brainstorm, which is a different and weaker activity. The premortem's power comes from the fact that the failures are generated privately, before social pressure has had a chance to operate on them. The room is full of people who know, with high confidence, why the project is in trouble. They will not say it out loud. They will, sometimes, write it on paper.
The writing has to be on paper, by the way, not on a shared screen. Shared screens are visible to the person sitting next to you. Visible to the person sitting next to you means visible to your boss. We have had clients insist that their digital whiteboarding tool is "anonymous." It rarely is in practice. The senior people know whose handwriting is whose, who tends to use which phrasing, who arrived late and submitted late. If you want unfiltered failures, you need actual unfiltered conditions. A piece of paper, folded once, collected by the facilitator, is sufficient. Anything more sophisticated is, in practice, less anonymous.
Failure mode three: sanitizing failures into safe categories before they're read
This one is subtle and possibly the most damaging. The facilitator collects the paper. The facilitator takes a break to "synthesize" the inputs. The facilitator returns with a slide that consolidates the seventeen individual failures into four headline themes: Market Dynamics, Execution Risk, Talent, Macroeconomic Conditions.
What was on the paper, before synthesis, was specific. "The head of product overrules the data team and we ship a feature we already know nobody wants." "Our second AE hire is going to ramp at the median of our last four, which is six months below the plan, and the plan does not survive that." "We don't actually have a decision-making body that can resolve the China / India trade-off and we'll spend the year not deciding."
After synthesis, what's left is talent and execution risk. The specifics — the things the team could actually do something about — have been scrubbed in the name of organizational palatability. The facilitator has done this, often unconsciously, to make the readout less awkward for the senior people in the room. The exact thing the technique was designed to surface has been quietly removed before anyone got to look at it.
The most useful sentence in a premortem is always the one someone almost didn't write down. The job of the facilitator is to make sure that sentence is read out, in the words it was written, before any of the polite category labels touch it.
The fix is to read the papers, unfiltered, in the words they were written. Anonymously. Out loud. The room sits with the specifics. The facilitator does not editorialize, does not group, does not relabel. After the reading, then — and only then — the room may cluster, vote, prioritize. The synthesis happens after the unfiltered specifics have landed. Not before.
Failure mode four: skipping the kill-criteria step
Most premortems we audit end at the risk list. The room generated the failures. The failures were discussed. Lessons were taken. Everyone files out feeling sober and prepared.
This is the cocktail-party version of a premortem. The operational version requires one more step, and most facilitators omit it: for each major identified failure, the room writes a kill criterion. What would we see, by when, that would tell us this failure is materializing? At what level of evidence do we change course?
This is the entire operational point of the exercise. Premortems are not about feeling pessimistic for an hour. They are about converting the eighteen-month death-by-a-thousand-meetings into a small number of explicit, scheduled checkpoints, agreed before the project starts, when the room is at its most calibrated. The team that wrote the kill criteria in month zero is the team that, in month nine, will use them. The team that didn't, won't — they will rationalize the bad signal, reframe the milestone, and ride the project into the wall.
A kill criterion is a sentence with a verb and a date. "If by Q3 our second AE has not closed at least one deal of $200k or more, we cut the regional plan in half." "If the integration committee has not produced a written decision-rights matrix by end of month four, we escalate to the CEO and reconsider the integration model." These are not aspirations. They are pre-commitments. They are the entire reason the meeting was worth having.
Failure mode five: not assigning owners
The last failure is the easiest to fix and the most often overlooked. Each surviving risk needs a named owner — not a function, not a team, a person — whose job is to watch the leading indicators, report the status at an agreed cadence, and raise the flag if the kill criterion is approaching. "We" is not an owner. "The product org" is not an owner. "Sarah" is an owner.
The risk that does not have an owner is the risk that, when it materializes, surprises everyone. We have walked into post-mortems where the failure mode that ended the project was the third item on the original premortem list, eighteen months earlier. Nobody had owned it. Nobody had watched it. Everyone, on the way out of the original meeting, had assumed someone else was on it. Nobody was.
What good facilitation actually looks like
A premortem run well — and we have, between us, run hundreds of these — has a recognizable shape. It takes about ninety minutes. Paper, not screens. Anonymous, not attributed. The senior person writes first, not speaks first. The papers are read aloud unfiltered before any synthesis happens. Each surviving risk gets a kill criterion with a date, in writing. Each criterion gets a named owner. The output document is one page, signed by everyone in the room, filed in the same place as the deck.
That document is not a CYA exercise. It is the single most valuable artifact a leadership team can put in the file alongside the original case. If the project succeeds, the document is a record of what the team correctly worried about and managed. If the project fails, the document is a record of where the team was right and where it was wrong — and that record is the only mechanism we know that turns one failed project into a smarter next one.
The technique was correctly described by Klein decades ago. The literature is reasonably mature. The failure modes we have catalogued here are all, individually, well-known to the small community of people who run premortems for a living. The reason most premortems still fail in business settings is that the facilitation has been delegated to amateurs — usually the project manager, usually with the best intentions, usually without any of the mechanical training that separates a useful exercise from a piece of theater.
We do not think this is unfixable. We think it requires that facilitation be taken as seriously as the technique itself. A premortem is a small, structured intervention with a large, asymmetric payoff. Done well, it identifies the failure mode that kills the project before the project starts. Done badly, it produces a slide that says "Market Dynamics" and lets everyone sleep through the next year.
If your team has been running premortems and the failures they identify have stopped feeling specific — or if they never did — that is exactly the room we are usually called into. We are happy to facilitate one with you, or to train your team to run them on their own. The technique is free. The facilitation is the part that has to be earned.
Writes about decision quality at Bayeseon. Reach the team at hello@bayeseon.com.