The Post-Mortem Playbook: turn one expensive failure into a rule you never break again

Failure is the most expensive thing you will buy this year. You already paid for it — in cash, in calendar, in someone’s confidence. The only question left is whether you also take delivery of the thing it was supposed to teach you. Most teams do not. They hold a meeting, feel bad in a conference room for an hour, produce an action item that says “improve communication,” and repeat the identical failure inside eighteen months with different names attached.

A real post-mortem is not a feelings exercise and it is not an investigation. It is a manufacturing process: raw material in (what actually happened), finished good out (a small number of changed defaults). Ninety minutes, one page, three commitments. Here is the whole thing.

Part 1 — The trigger rule: what earns one

Post-mortem everything and you will post-mortem nothing, because the practice dies of exhaustion by week six. Write the trigger down once and let it decide for you. Something earns a post-mortem when it clears any one of these bars:

• Money. It cost more than 1% of annual revenue, or more than one month of your personal runway.
• Time. It burned more than two weeks of anyone’s calendar with nothing shippable at the end.
• Trust. A customer, an investor or a teammate now believes something about your reliability that you would have to work to undo.
• Surprise. Nobody saw it coming until it was already true. Surprise is the most valuable trigger on the list, because it means your instrumentation failed, not just your execution.
• Repeat. It rhymes with something that happened before. Second occurrence is mandatory, no exceptions, no matter how small.

Everything else gets a two-line note in a running doc and no meeting. Also worth saying: run one on a win once a quarter. Wins are where undiagnosed luck lives, and luck that you have mistaken for skill is the most expensive belief you can hold.

Part 2 — The 48-hour window

Hold it between 48 hours and seven days after the dust settles. Sooner and people are still defending; later and the timeline dissolves into vibes. Book 90 minutes, name a facilitator who did not own the work, and send this note verbatim:

“We’re running a post-mortem on [X] Thursday at 2. It is not an investigation and no one is on the hook. Before Wednesday 5pm, add your entries to the shared timeline doc: what you did, when you did it, and what you believed to be true at that moment. Facts and beliefs only — no conclusions yet, we do those together in the room.”

That last clause is the whole trick. Asking for beliefs alongside actions is what makes the meeting blameless without anyone having to say the word “blameless.”

Part 3 — Build the timeline before you build the story

One shared document, one row per event, four columns: timestamp / what happened / who knew / what we believed at the time. Everyone fills in their own rows asynchronously. No commentary, no adjectives, no “obviously.”

A real row looks like this: “March 14, 9:40am — approved the scope change by Slack thumbs-up. Only I knew. I believed the client had already signed off internally because they said ‘we’re aligned.’” That single row contains the entire failure: a decision made on an inference, recorded nowhere, visible to no one. You will find that most disasters are three or four such rows sitting quietly next to each other, and not one villain.

Then mark two things on the finished timeline before the meeting starts. The last honest moment — the point at which a cheap intervention would still have worked. And the detection gap — the distance between when it became true and when someone noticed. Those two markers are usually the actual subject of the meeting.

Part 4 — The only four questions

“Five whys” drills a single hole in a wall that has several. Ask these four instead, in this order, out loud, one at a time:

1. What did we believe that turned out to be false? Not what went wrong — what was believed. Every failure is a bad belief that got funded.
2. What made that belief reasonable at the time? If you cannot answer this one generously, you have not understood the failure yet, you have just found someone to be annoyed at. Reasonable-at-the-time is the definition of a systems problem.
3. What would have told us sooner? The signal that existed and was not watched, or the signal that did not exist and should. This is the question that produces the most durable fixes.
4. What did we do right that we should keep? Ask it even when it feels absurd. Something contained the damage; if you do not name it, the reorganization that follows will quietly delete it.

Two facilitator scripts to have loaded, because both moments will happen. When the room starts hunting a person: “Let’s assume everyone made the best call available with what they had in front of them. What did they have in front of them?” When someone starts pre-defending: “I’m not asking why you decided that — I’m asking what you could see when you decided it.”

Part 5 — The 90-minute agenda

• 0–10. Facilitator reads the timeline aloud, start to finish. Reading it out loud is not ceremony — it is how people discover which rows they had never seen.
• 10–20. Corrections and missing rows only. Still no analysis.
• 20–50. The four questions. Facilitator writes answers on the wall where everyone can see them.
• 50–70. Candidate changes. Generate freely, judge nothing yet.
• 70–85. Cut to three. Each survivor gets a named owner and a date. Then pick the one thing you will stop doing.
• 85–90. Read the three back. Whoever owns each one says “yes” out loud. Meeting ends.

Part 6 — The one-page output

If it is longer than a page, no one will read it in eight months when it matters. The format, copy it exactly:

• What happened — three sentences, no adjectives.
• What it cost — a number. Dollars, days, or the named account.
• The false belief — one sentence, in plain language.
• The detection gap — how long it was true before anyone knew.
• Three changes — each with an owner and a date, written as a changed default, not an intention.
• One stop-doing — the practice you are retiring.
• Keep — the thing that worked.

The difference between an intention and a changed default is the whole game. “Be more careful about scope” is an intention and it will decay in eleven days. “Any scope change over $2k requires a one-line email from the client before work starts — template lives here, in the proposal doc, Priya owns it, live August 15” is a default. Defaults survive your motivation, which is the point, because your motivation is not invited. If the fix requires someone to remember to be virtuous, it is not a fix. A good default is one where doing it wrong takes more effort than doing it right — a required field, a checklist item, a calendar block, a two-signature rule. The same logic drives the scope playbook and the handoff playbook: encode it once, stop relying on memory.

Part 7 — The six failure modes

• The trial. One person answers 80% of the questions. Fix: facilitator who did not own the work, and the timeline read aloud before anyone speaks.
• The fog. Everything is “a communication issue.” That phrase is a fog machine. Push until you have a specific message that a specific person did not receive at a specific hour.
• The twelve-item list. Twelve action items is zero action items. Three is the maximum a team actually installs. Cutting nine good ideas is the job.
• The orphan. Changes with no owner or no date. If nobody said “yes” out loud, it is not a commitment, it is a mood.
• The vanish. The page goes in a folder nobody opens. Fix: read the three commitments at the top of the next monthly review, and mark each one shipped, dropped or slipping. Ninety seconds.
• The over-correction. One bad outcome triggers a rule that taxes every future project. Ask: would this change have prevented it, and what does it cost when things go right? If the second answer is bigger, do not ship it.

Part 8 — The solo version, 25 minutes

No team? The mechanics survive intact, and honestly it is easier, because the only person you have to stop protecting is yourself. Ten minutes: write the timeline from calendar, sent mail and bank statements — artifacts, not memory. Ten minutes: answer the four questions in writing, and be specific about what made the false belief reasonable. Five minutes: three changed defaults, one stop-doing, dates on all of them, and a calendar reminder 30 days out titled “did I actually do these.”

Then the part that makes it stick: keep one running file called Rules I paid for, one line per lesson, with the price next to it. “Never start design work before the deposit clears — $8,400.” “Never hire someone I have not seen handle a disagreement — 4 months.” Re-read it before every new project. Twenty lines in, that file is the most valuable document you own, and it is the closest thing to a cheat code that exists in professional life: the ability to buy a lesson once instead of renting it repeatedly.

The objection clinic

• “It will turn into a blame session.” It will if you start with opinions. The written timeline plus the reasonable-at-the-time question makes blame structurally awkward — you cannot hunt a person while the room is busy establishing that their call made sense.
• “We already know what went wrong.” You know what went wrong last. The four questions find what went wrong first, which is usually three weeks earlier and completely different. Run the timeline anyway; the cheapest version costs 25 minutes.
• “Everyone’s morale is already low.” Unexamined failure is what tanks morale — people know something broke and are privately assigning blame with no data. A structured process is a relief. What is corrosive is the silence.
• “We don’t have 90 minutes.” You had six weeks for the failure. And if the trigger rule is doing its job, this is happening a handful of times a year, not weekly.
• “The cause was outside our control.” Fine — then the question is not the cause, it is the detection gap and the response. You cannot control the market. You can control how many days it takes you to notice it moved.

Your first one, this week

Pick the failure you are still privately annoyed about — that irritation is a reliable instrument, it means the lesson has not been extracted yet. Today: open a doc, four columns, and put in your own rows from calendar and email. Tomorrow: if others were involved, send the note from Part 2 and ask for their rows by end of day. Thursday: 90 minutes, the agenda above, or 25 minutes alone. Friday: the one page goes somewhere findable, the three owners have dates in their calendars, and the stop-doing gets deleted from wherever it lives — the template, the checklist, the recurring meeting.

Do it four times a year and you compound something rarer than skill: an organization, or a career, that only makes each mistake once. Everyone else is paying full price, over and over, for lessons they have already bought.