Post-mortems without blame
A retro where everyone is polite is a retro where nobody learned anything. Getting the truth into the room is a design problem, not a courage problem.
Two versions of the same meeting.
In the first, the project just wrapped, it went sideways in week four, and everyone knows exactly why. The retro produces a slide that says “communication could improve” and three action items nobody will read again. Polite nods. Meeting ends early. The same failure is now scheduled to reappear in the next project, on time and under budget.
Version two: somebody says the true thing. The handoff broke because two people each believed the other owned it, and the file format made the mistake invisible for nine days. Uncomfortable for a minute. Then the room gets to work on the actual problem, which it finally knows about.
The distance between those two meetings isn’t courage. It’s design.
Why people lie in retros
They lie because lying is rational. If naming the real cause might mark a colleague, or themselves, people will sand the truth down to “communication could improve” every single time, and they’re right to, because they’ve correctly read what the room rewards. Blame is a tax on truth. Raise the tax high enough and the supply drops to zero.
Fields where mistakes cost lives figured this out generations ago. Aviation and medicine both learned, painfully, that punishing whoever reports an error mostly teaches people to stop reporting errors. The error rate doesn’t drop. What drops is the reporting rate, which is worse, because now the failure repeats in the dark.
Agencies aren’t landing planes. But the mechanism is identical at any scale: a retro’s output is limited by what people are willing to say in it, and what people are willing to say is set by what happened to the last person who told the truth.
Systems over culprits
The person who missed the handoff is not the cause of the missed handoff. They’re the last domino, the one that happened to be standing where the failure came out. The cause lives upstream, in a process where one unread message could silently orphan a deliverable for nine days. Fire the person, hire a replacement, and the trap sits there patiently, waiting for the new hire to be tired on the wrong Tuesday.
Culprits are cheap to find, satisfying to name, and useless to fix. Systems are the opposite on all three counts.
The question that does the work
One phrase carries more weight in our retros than everything else combined. For each failure on the timeline, we ask:
What would have made this impossible to get wrong?
Not “who should have caught it.” And not “how do we catch it next time,” either, which quietly assumes the failure gets another chance. Impossible is the standard, and reaching for it changes the kind of answers a room produces. A person forgot to update the tracker: what if the tracker updated itself? Two people each thought the other owned it: what if ownership were a field instead of a vibe? The deadline surprised everyone: what if the schedule lived where the work lives, not in a PDF from March?
You won’t always reach impossible. Sometimes the honest answer is a checklist, which merely makes the mistake embarrassing instead of easy. Fine. The point of the question isn’t perfection, it’s direction: it aims the room at the mechanism instead of the operator, and the shift is audible: people stop defending themselves and start engineering, often within the same sentence.
And here’s the test for whether you’ve actually answered it. If the fix is “people will be more careful,” you haven’t found the fix. Careful is what people were already trying to be.
Mechanics that keep it honest
A few rules we hold to.
- Timeline first, interpretations second: the room agrees on what happened, in order, before anyone argues about why.
- Whoever holds the most power speaks first and confesses first, because nothing licenses honesty faster than the project lead opening with their own worst call.
- Retros stay severed from performance reviews, structurally, not as a pinky promise.
- The meeting ends with system changes that have owners and dates, never with vows to try harder, because a vow is just blame with a delay on it.
None of this makes retros comfortable. Truth arriving in a conference room rarely is. It makes them safe, which is different, and better, and the only condition under which the truth shows up at all.
The culprit walks out of a good retro still employed. It’s the mistake that doesn’t.
