Pre-mortem: the template, with worked examples

Cerno is a private AI boardroom that puts consequential decisions through a structured process — advisor challenge, a recorded vote, risk and confidence scores, and an Executive Decision Brief. The pre-mortem is one of the lenses it applies; here's how the technique works on its own, how to run it alone, and the ways it goes wrong.

What is a pre-mortem?

Coined by the psychologist Gary Klein, a pre-mortem is the opposite of a post-mortem. Instead of asking "why did this fail?" after the fact, you ask it before you commit: "It's a year from now and this decision was a disaster. What went wrong?"

The framing matters. Asked to critique a plan directly, people hedge — nobody wants to be the naysayer, least of all when the plan's owner is in the room. Asked to explain a failure that has (hypothetically) already happened, the same people become fluent and specific. You get the real risks, not the polite ones.

Why it works

Three things happen when you assume the failure rather than merely allow for it.

It gives permission. "This might not work" is an opinion, and offering it makes you the pessimist. "Here is why it failed" is an assignment, and completing it makes you helpful. The social cost of raising a doubt drops to nothing.

It turns unease into mechanism. Most people know, vaguely, that a plan has a soft spot. A pre-mortem forces the vague feeling into a sentence with a cause in it: not "the launch might go badly" but "we launched before the payment flow was tested and the first fifty customers hit a bug." A risk with a mechanism can be mitigated. A feeling cannot.

Hindsight is easier than foresight. Research on what psychologists call prospective hindsight — imagining an outcome as having already happened — finds that people generate noticeably more, and more specific, reasons for it than when they are asked whether it might happen. Certainty, even pretend certainty, unlocks detail.

When should you use it?

Use a pre-mortem before any decision that is consequential and hard to reverse: a product launch, a key hire, a market entry, a large capital commitment, a long lease. The test is simple — if an unspotted failure mode would be expensive or embarrassing, spend twenty minutes here first.

Skip it for small, reversible calls. If you can undo the decision cheaply, you don't need to imagine its failure; you need to make it and find out.

How to run a pre-mortem in 5 steps

1. ›Set the scene. State the decision as if it's already been made and has failed: "We shipped, and six months on it's clearly the wrong call." Be concrete about the timeframe and about how bad — "clearly wrong" is better than "not ideal".
2. ›Everyone writes independently. Each person lists the reasons for the failure — alone, before hearing anyone else, so no one anchors on the loudest voice or the most senior one. Two minutes, as many reasons as they can.
3. ›Collect every reason. Go round the room, one reason each, until the lists are empty. Pool them without debate. The goal is breadth, not consensus yet.
4. ›Cluster and rank. Group the failure modes that are really the same thing, then rank by likelihood × impact. The top three or four are your real risks.
5. ›Turn risks into actions. For each top risk, decide one of three things: a mitigation (something you change now), a tripwire (an early-warning signal you'll watch for), or a stop condition (a fact that, if it turns out true, means you don't proceed).

Three prompts get better answers at step 2 than "what went wrong?" alone:

• ›What specifically broke? Not "the market" — what happened, in what order.
• ›When did we first see a sign of it? Usually earlier than the failure, and that's your tripwire.
• ›What did we know at the start and ignore? This one produces the uncomfortable, useful answers.

Worked example

You're weighing whether to open a second site. A pre-mortem surfaces the failure modes, and the brief records what to do about them:

Notice what the brief did with the pre-mortem's output. The top failure mode — demand doesn't follow — became a stop condition (the demand test before signing) and a tripwire (weekly bookings against forecast). The second — the site runs with nobody running it — became a gate. The pre-mortem didn't say no; it said "not like this".

Running one on your own

Most small businesses don't have a room to fill, and the standard advice quietly assumes one. A pre-mortem works alone, with two changes.

Write before you judge. The whole technique depends on separating generating reasons from evaluating them. On your own, the evaluation creeps in instantly — you think of a reason and dismiss it in the same breath. So set a timer for ten minutes, write every reason for the failure you can think of, and do not cross anything out. Aim for ten. The seventh and eighth are usually the ones you'd never have said aloud.

Come back tomorrow. A second person's job in a pre-mortem is partly to not share your optimism. Sleep does some of that work. Rank the list the next morning, when the plan is less exciting and the list is just a list.

If you want the second voice without waiting a day, that's what the Risk Officer advisor in Cerno is for: it assumes the plan has failed and enumerates why, without any stake in the plan going ahead.

Common mistakes

Generic risks. "The economy could turn." "A competitor might launch something." These are true of every plan and therefore tell you nothing about this one. If a reason on the list would appear in any pre-mortem, it isn't doing any work. Push for the mechanism specific to this decision.

Running it after the decision. A pre-mortem on a plan you've already committed to produces reassurance, not risks. The people in the room know the answer is "go", and they list failure modes that are safely mitigable. Run it while "don't" is still a real option.

Letting the plan's owner chair it. Whoever proposed the plan will, without meaning to, steer the discussion towards the risks they've already thought about. Someone else runs the room; the owner writes their list like everyone else.

Stopping at the list. A pre-mortem that ends with a ranked list of risks has done half the job. Each top risk needs a mitigation, a tripwire or a stop condition — and the tripwires need to go in the calendar, or nobody will watch for them.

Treating it as a veto. The output is almost never "abandon the plan". It is usually "proceed, with these three changes and these two things to watch". A pre-mortem that always says no is being run by the wrong person.

Which Cerno advisor uses this?