There is a meeting that does not get scheduled often enough. It takes about an hour. It requires no consultant, no software, no model, and no data the room does not already have. It tends to identify, with eerie consistency, exactly the failure mode that does in fact kill the project eighteen months later. And most teams, having heard of it, never actually do it.
The premortem, in its simplest form: gather the people who will execute the decision. Tell them to assume it is two years from now and the project has failed badly. Ask them, individually and in writing, why. Then read the answers out loud.
The technique is decades old. Gary Klein popularized it. It is taught in every decision-science course we know of. It is the kind of thing that, if you described it cold to a CEO, would sound almost insultingly obvious. And yet — in the engagements where we run them, more than half the time, the team has never done one before for a decision of comparable size.
Why people skip it
The official reason is always time. The premortem doesn't take time. An hour, run well, is enough. Two hours is generous. The actual reasons are different, and worth naming.
The first is commitment. By the time a major decision is being framed, the team has invested months getting to the point where it could be framed at all. Asking them to imagine, in detail, that the decision was wrong feels like asking them to undo their own work. It is uncomfortable in a specific way: the discomfort of a sunk-cost fallacy made conscious. Most teams avoid it without admitting they are avoiding it.
The second is hierarchy. The premortem only works if junior members can write, on paper, the things they would never say to the SVP's face. In most companies they can't, and they won't, and the SVP knows it. The technique requires actual psychological safety, not the kind that appears in town halls. If your team is not willing to write, anonymously, "the head of product overrules the data and we ship a feature nobody asked for," your premortem will produce a list of weather, regulation, and the macro environment. Which is to say, a list of things nobody in the room can do anything about.
The third is optics. A premortem produces a document. The document says, in writing, that the people who recommended the decision could see how it might fail. If the decision then fails, that document is evidence. Better, the reasoning goes, not to produce it at all.
This is exactly backwards. The document is evidence that the team did its job. A board that punishes a written list of risks is a board that has trained its executives to be ignorant on the record. The premortem document, in our experience, is one of the single most valuable artifacts a leadership team can put in the file — both for the decision in front of them and for the audit, two years later, of how they decided.
What a good one produces
A premortem done well produces three things.
It produces a risk list ranked by named consequence. Not "market conditions" but: "the regional team we just hired turns over, and we lose six months." Not "execution risk" but: "the integration committee meets monthly, the team needs weekly decisions, and we burn the year arguing about decision rights." Specifics are what make the list usable. Generalities are what make it a CYA exercise.
The most useful sentence in a premortem is always the one someone almost didn't write down. If your premortem ends without anyone being embarrassed, it didn't work.
It produces a set of kill criteria. For each major risk, what would we see, by when, that would tell us this decision is going wrong? The act of stating these in advance is almost magical: it converts the eighteen-month death-by-a-thousand-meetings into a clear, scheduled decision point. The team that wrote the criteria is the team that, six months in, will actually use them. The team that didn't, won't.
It produces a named owner per risk. Not "we'll keep an eye on it." A person, a metric, a check-in cadence. The risk that does not have an owner is the risk that, when it materializes, surprises everyone.
What it doesn't do
The premortem is not a forecast. It will not tell you whether the decision is right; it will not produce a probability of success; it will not replace any of the harder work of framing the decision properly. What it does, reliably, is reveal the part of the analysis the team has been working hard not to look at. Sometimes that part is decisive. Sometimes it is a footnote. Either way, the team is now operating on a more honest picture.
The cost of running one is one hour and one piece of organizational courage. The cost of not running one — the cost of the meetings, eighteen months later, that try to work out how something so obvious in hindsight was missed in foresight — is much, much higher. We have never seen a leadership team that ran a real premortem and regretted the hour. We have seen many that wish, retrospectively, they had given it.
If you have a decision in the next quarter where the failure mode is, in fact, predictable, and the room around you is too aligned to be questioned — that is the room that most needs the meeting. We are happy to run one with you, or to teach your team to run them on their own. The technique is, again, free. Most things worth doing aren't.
Writes about decision quality at Bayeseon. Reach the team at hello@bayeseon.com.