How to write a failure post-mortem

A post-mortem is worth writing for one reason: the next attempt should fail for a new reason, not the same one. Everything else people do with these documents, the tidy narrative, the lessons for LinkedIn, is decoration.

Write it within two weeks

Memory reorganises fast. Within a month you will have a clean story where the cause was the market or the co-founder or the timing. The version written while you are still annoyed contains the dates, the moment you ignored something, and the number you did not want to look at. Those are the parts that help.

The five questions

1. What did you believe that was false? Write the belief as a sentence someone could have disagreed with at the time. "People would pay for this monthly" is a belief. "We had no product market fit" is a label.

2. When was the first evidence against it? Find the date. There is almost always a moment months before the end where something told you, and you explained it away. Write down what you told yourself.

3. What actually ran out? Money, customers, time, or your own will. These are four different failures and they have four different fixes. Founders default to blaming money because it is the least personal answer.

4. What did you do that you would do again? There is always something. Skipping this question makes the document useless, because next time you will throw out the good parts along with the bad.

5. What would early success look like next time? One measurable thing, visible in the first month. If you cannot name it, you are not ready to start again.

Three traps

The villain. If the document has one person or one event as the cause, it is a story, not an analysis. The co-founder who left is a fact. Why the company could not survive one person leaving is the finding.

The lesson list. "Hire slowly, fire fast" and the rest are proverbs. They are true for everyone and useful to nobody. Replace each proverb with the specific thing you will do differently, with a date attached.

The audience. Write it for yourself first. A version for the internet can come later, and it will be a different document with less in it.

The template

PROJECT: [name] · attempt #[N]
RAN FROM: [start] to [end]

WHAT I BELIEVED
[The sentence someone could have disagreed with.]

FIRST EVIDENCE AGAINST IT
[Date.] [What happened.] [What I told myself instead.]

WHAT RAN OUT
[Money / customers / time / will.] [The number on the day I stopped.]

WHAT I WOULD DO AGAIN
[One or two things that worked.]

WHAT I WILL DO DIFFERENTLY
[Specific action.] [By when.]

EARLY SUCCESS NEXT TIME LOOKS LIKE
[One measurable thing, visible within 30 days.]

COUNT
This was attempt [N]. The next one is [N+1].

Where to put it

Somewhere another person will read it. A document nobody sees gets written to flatter you, and you will not notice you are doing it. In our server people post theirs in the failures channel and someone two or three attempts further along answers. That reply is usually where the real finding comes from, because they have made the same mistake and can name it faster than you can.

Applying takes five minutes and one of us reads every one.