The Pre-Mortem: How to Kill Your Startup on Paper Before It Dies for Real

Every founder can tell you why their startup will win. Almost none can tell you, in specific and unflattering detail, how it will die. That asymmetry is the whole problem.

Your deck is a monument to optimism. It has to be — you can’t recruit, raise or ship on a foundation of doubt. But the same optimism that gets the thing off the ground is exactly what blinds you to the two or three failure modes that will actually end it. You are structurally incentivized not to look.

A pre-mortem is the fix. It’s a technique borrowed from Gary Klein and popularized by Daniel Kahneman, and the mechanics are almost embarrassingly simple: you assume the failure has already happened, then you explain it. Not “what could go wrong” — that’s a risk register, and it’s toothless. The move is to time-travel to the funeral and write the obituary. It sounds like a party trick. It is one of the highest-leverage hours you will spend before you build anything.

Why the framing does the work

The magic isn’t in listing risks. You can list risks all day and feel productive while learning nothing. The magic is in the grammatical shift from could to did.

Klein’s original research found that imagining an outcome as certain — “the project has failed, it’s a fact” — increased people’s ability to correctly identify reasons for future outcomes by about 30%. This is called prospective hindsight, and it works because certainty unlocks candor. When failure is hypothetical, saying it out loud feels disloyal, almost like sabotage. When failure is a given and you’re just doing forensics, the same observation becomes helpful, even obligatory.

“What could go wrong?” invites a shrug and a risk register. “It’s three years later and we’re dead — what killed us?” invites a confession.

That’s the reframe most people miss. They run something they call a pre-mortem, but they’re actually running a risk-assessment meeting with a spookier name. They generate a tidy list of manageable-sounding risks — “competition,” “hiring,” “runway” — assign each a color, and move on. Nobody’s pulse changed. That’s not a pre-mortem. That’s theater.

How to run one that bites

A real pre-mortem produces one or two specific, named, plausible causes of death that you didn’t want to look at — and then changes a decision. Here’s the sequence.

1. Fix the timeframe and declare death. Pick a horizon — 18 months, 3 years — and state it as fact: “It is 2026. The company is dead / has shut down / raised no follow-on and wound down.” No hedging. The failure is settled; you’re here to explain it.
2. Write the obituary independently, then pool. If you have a team, everyone writes their own causes of death alone first, before anyone speaks. Groups converge on the loudest voice’s story; silent writing preserves the outlier who saw the real thing. Solo founder? Force the same discipline — write three distinct death scenarios before you let yourself react to any of them.
3. Push past the first, flattering answer. The first cause everyone reaches for is external and blameless: “the market wasn’t ready,” “a competitor out-raised us.” Those are alibis. Keep asking “and what let that kill us?” until you hit something internal and specific: you never found a repeatable channel, the founding team split over direction, the product solved a problem people had but wouldn’t pay to fix.
4. Build the collapse timeline. A cause is a bumper sticker. A timeline is a diagnosis. Sequence it quarter by quarter: what looked fine in Q1, the leading indicator you rationalized in Q3, the point of no return in Q6. Failures rarely arrive as a single event — they compound. Seeing the sequence tells you which early signal to instrument now.
5. Convert each cause into a decision. This is where 90% of pre-mortems die. A list you file away is worthless. For every serious cause, name one thing you’ll do differently this month — a metric you’ll watch, an experiment you’ll run, a conversation you’ll have — or explicitly decide to accept the risk. No decision, no point.

The common mistakes

• Softening the premise. “What if things get hard?” is not the exercise. The failure must be stated as complete and certain, or the candor never shows up.
• Stopping at external causes. “The market changed” is where lazy pre-mortems end and honest ones begin. Almost every external death had an internal enabler.
• Confusing volume with insight. Twenty generic risks are worse than three specific ones — the noise buries the two that matter.
• Doing it once. Your failure modes at pre-seed are not your failure modes at Series A. Re-run it whenever the shape of the risk changes.
• No owner, no action. A cause without a corresponding decision is a diary entry, not a defense.

A worked mini-example

Say you’re building an AI tool that drafts contracts for solo lawyers. The optimistic deck says: huge underserved market, clear pain, obvious willingness to pay.

Now run the pre-mortem. It’s three years out and you’ve shut down. Why?

The lazy answer: “Incumbents added AI features and out-marketed us.” Fine — but what let that kill you? Push. The real timeline emerges: Q2 — signups were healthy but activation was quietly dismal; lawyers tried it once and didn’t return because they didn’t trust an unreviewed draft. Q4 — you papered over retention with paid acquisition, so the burn looked like growth. Q6 — the paid channel got expensive, retention was still broken underneath, and there was no organic engine to fall back on.

The real cause of death wasn’t the incumbent. It was that you never solved trust-driven activation, and you hid the symptom with ad spend. That reframes what you do this month: instrument week-two retention now, and treat any activation problem as existential rather than a growth-marketing patch. That’s a pre-mortem earning its keep.

The problem with doing this by hand

The pre-mortem is nearly free and almost nobody runs one properly. The reasons are predictable. You’re the founder — you’re the least able person to imagine your own failure vividly, because your entire psychology is organized around the opposite. Left to your own devices, you’ll generate the flattering, external, blameless causes and quietly skip the ones that implicate a decision you’ve already made.

Doing it with a team helps, but coordinating an honest one is its own tax: someone has to facilitate, people hedge in front of the boss, and the quarter-by-quarter timeline — the part that’s actually diagnostic — is tedious enough that most groups never build it. So the exercise gets skipped, or worse, faked: a five-minute risk list that lets everyone feel rigorous without anyone feeling uncomfortable. The discomfort was the point.

Where Risk Matrix comes in

This is the narrow, useful thing Risk Matrix does. You give it your startup; it assumes you’ve already failed three years out and writes the autopsy — five distinct failure scenarios plus a quarter-by-quarter collapse timeline.

The value is structural. Because the tool starts from “you’re already dead,” it has no incentive to protect your ego — it skips straight past the flattering external alibis to the specific, plausible causes you’d avoid on your own. Five scenarios force breadth past the one obvious risk you fixate on. And the quarter-by-quarter timeline is exactly the tedious, diagnostic part most manual pre-mortems never bother to build — the sequence that tells you which early signal to instrument today. It doesn’t replace your judgment about which deaths to take seriously. It gets a rigorous, unflinching first draft on the table in minutes instead of never.

Try Risk Matrix →

It’s not the only way

A pre-mortem is a technique, not a product, and there are several honest ways to run one. Pick by what you’re short on — time, objectivity or facilitation skill.

The bottom line

Optimism builds the company; a pre-mortem keeps it honest. The founders who survive aren’t the ones who never see failure coming — they’re the ones who forced themselves to write the obituary early, spotted the one internal cause hiding behind the flattering external one, and changed a decision because of it. Do it by hand if you have a sharp team and an hour of real candor. Use a tool if you’d otherwise skip it or fake it. Just don’t confuse a color-coded risk list with actually looking death in the face.

Try Risk Matrix · Get The Brief