Running a Post-Mortem Without Blame That Still Changes Something
How to run a review after a failure that finds real causes without scapegoating anyone or letting accountability disappear, using Amy Edmondson, W. Edwards Deming, and Gene Kranz.
The post-mortem is on the calendar for Friday. Everyone has already decided what they will not say. The only open question is whether the meeting produces a document called "lessons learned" that nobody opens again.
What goes wrong with the usual post-mortem?
Three things, in roughly this order. People protect themselves, so the account is sanitized. The group settles on a proximate cause, usually a person or a step, and calls it root cause. And the actions are vague enough that nobody can be late on them.
Ask an assistant for "a post-mortem template" and you get headings: summary, timeline, root cause, lessons. The headings are fine. They cannot supply candor, and they do not stop the review from ending at the first plausible answer.
Each failure mode has a documented response.
What if people are protecting themselves?
Amy Edmondson's The Fearless Organization (2019) describes psychological safety as the belief that speaking up carries no punishment for honest mistakes or questions. She describes three leader tasks: set the stage by framing the work as a learning problem, invite participation through genuine questions, and respond productively when people speak.
Her later book Right Kind of Wrong (2023) distinguishes failure types. Basic failures come from error in known territory and are largely preventable. Complex failures arise from a combination of factors. Intelligent failures happen in new territory, with a sound hypothesis and as little cost as practical.
Applied: Open the review by naming which kind of failure you think it was, then invite correction. Go first on your own part in it. Ask "what did you know at the time?" instead of "why did you do that?"
"Here is what happened: [short description]. Using the distinction between basic, complex, and intelligent failures, help me decide which kind this most likely was and how deep the review should go. Then draft three opening sentences I could use to frame the meeting as a learning exercise, including one acknowledging my own role."
Why it works: classifying the failure keeps effort proportional and signals that the purpose is learning.
What if the "root cause" is just the nearest person?
W. Edwards Deming's Out of the Crisis (1982) argues that the great majority of problems come from the system, which is management's responsibility, rather than from the individual worker. His often-cited estimate puts it at around 94 percent system and 6 percent special causes. He also named driving out fear as one of his Fourteen Points, because people who are afraid hide the information improvement depends on.
Applied: When a cause names a person, ask what made it easy to do or hard to catch. Missing checklist, unclear ownership, rushed schedule, tool that permits the error. Keep asking why until the answer is something you can change in a process.
"The proposed cause of this failure is "[person or step]". Ask me five "why" questions in sequence, each digging into the conditions that made this possible: process, incentives, tooling, information, and schedule. Then summarize the system-level causes I could change and mark which are within our control."
Why it works: repeating the question moves attention from the person to the conditions you can alter.
What if blameless slides into nobody owning anything?
Gene Kranz's Failure Is Not an Option (2000) describes the Monday after the Apollo 1 fire in 1967, when he addressed his flight controllers. He did not hunt for a culprit. He told them they had become too focused on the schedule and had overlooked problems they saw in their own work, and that from then on the team would be defined by two words: tough and competent. Tough meant accountable for what you do and fail to do. Competent meant taking nothing for granted.
Applied: Pair every cause with an owner and a date. Blameless describes how you talk about the past. Accountable describes who changes the future.
"These are the findings from our post-mortem: [list]. Convert each into a specific action with a named owner, a date, and a way we will know it is done. Flag any action that is too vague to be late, and suggest a check-in date to review progress."
Why it works: owners and dates convert insight into commitments someone can fail to keep, which is the point.
Which approach do you lean on?
| Approach | From | When it fits | Watch out for |
|---|---|---|---|
| Safety and failure types | Amy Edmondson, The Fearless Organization and Right Kind of Wrong | People are guarded and the review feels staged | Treating safety as a speech rather than behavior over time |
| System over person | W. Edwards Deming, Out of the Crisis | The proposed cause is an individual and the fix is "be more careful" | Using "the system" to excuse repeated individual negligence |
| Ownership without scapegoats | Gene Kranz, Failure Is Not an Option | The group is relieved of blame and drifting toward no action | Mistaking toughness for harshness toward people |
What can't a post-mortem do?
It cannot undo the failure, and it cannot create trust where none exists. If people have been punished for honesty before, one good meeting will not change the memory. It also will not decide whether the underlying bet was sound. Sometimes a decision was reasonable and the outcome unlucky, and the honest review says so. A structured separation of decision quality from outcome luck is a different exercise.
Amy Edmondson is in the People & Culture category, W. Edwards Deming in Operator & Systems, and Gene Kranz in Explorer & Adventurer. For the originals, read The Fearless Organization, Right Kind of Wrong, Out of the Crisis, and Failure Is Not an Option. To separate decision quality from outcome luck after the fact, What Actually Happened is a $79 tool built for that.
All the copy-paste prompts from this guide, and the rest of the Situations series, live at /prompts too, free to copy without reading the guide first.
Frequently asked questions
Does blameless mean nobody is accountable?
No. Edmondson and Kranz both separate blame, which is about punishing a person, from accountability, which is about owning decisions and their consequences. Kranz's Apollo 1 response was a demand for ownership, not a search for a villain. A blameless review asks what happened and who owns the fix. It does not pretend nobody made choices. Say this explicitly at the start of the review, because people often hear blameless as permission to avoid hard facts.
What if one person really did make a serious mistake?
Then the review should say what they did, what information they had, and why the system allowed it to matter. Deming's view is that most failures trace to the system, but he did not claim individuals never err. If the pattern is repeated negligence, that is a performance matter handled separately from the learning review. Document the individual contribution factually, ask what would have prevented it, and keep any performance conversation in a separate, private setting.
How soon after a failure should we hold the review?
Once the immediate crisis is contained and people have slept, usually within days rather than weeks. Memory fades and stories harden. Too soon and people are defensive or exhausted. Pick a date, write the timeline together beforehand, and keep the meeting about what to change. A short written timeline built from messages and logs before the meeting saves time and reduces arguments about what actually happened.
How do I stop people from only saying what is safe?
Edmondson's work points to the leader's own behavior. Frame the review as a learning problem, ask open questions, and respond to uncomfortable facts with appreciation. If the first honest answer gets punished, the second one never arrives. Starting with your own contribution to the failure sets the tone. Model the behavior yourself by describing what you did not know, and by thanking whoever raises the awkward point.
Are all failures worth a deep review?
No. Edmondson's Right Kind of Wrong separates basic, complex, and intelligent failures. Basic failures in familiar territory deserve a quick fix. Complex failures with many causes reward a deeper review. Intelligent failures in genuinely new territory deserve celebration for what they taught, not an inquest. Matching effort to the kind of failure also protects your team from review fatigue, which quietly kills the habit.
Can AI facilitate the review?
It can help build a timeline from messages, group causes by type, and draft action items with owners. It cannot judge candor in the room or tell when someone is holding back. Use it to organize the findings and keep the human conversation for the human parts. Ask it to challenge each proposed cause and list evidence for and against, which helps expose assumptions presented as facts.
Written by Gareth Hoyle. Last updated 8 October 2026. Part of the authority.md guides library.
More guides.
Apologizing to a Customer When You Got It Wrong
How to write an apology a customer believes, using the Trust Triangle from Frances Frei and Anne Morriss, the Trust Equation from Maister, Green and Galford, and Robert Cialdini's commitment principle used ethically.
Asking for a Raise Without Turning It Into a Standoff
How to run a pay conversation as a problem you and your manager solve together, using three documented negotiation approaches from Fisher and Ury, Chris Voss, and Deepak Malhotra.
Client Scope Creep: Before, During, and After It Happens
How to stop a client's small extra requests from becoming unpaid work, using documented approaches from Blair Enns, William Ury, and Maister, Green and Galford's Trusted Advisor.