What practices make a company's postmortem process actually effective rather than symbolic?
6 recorded positions from 3 people, first said May 11, 2022. They do not agree — the readings below are what each one actually argued.
Also on the record
Tony Fadell · May 11, 2022
A good postmortem covers what you got right as well as wrong and separates macro environmental causes from internal process failures
You can't blame a business for a macro event like the pandemic, but failing to pivot within that environment is your own process problem; splitting the two lets you spot and avoid the failure modes next time
42:08 Good postmortems separate macro environmental causes from internal process failures and credit successes too
Max Levchin · Feb 5, 2025
A postmortem culture requires strict segregation of each incident's data into its own channel and document from the outset
Good postmortem cultures generate huge volumes of data, much of it irrelevant, and it's very hard to peel apart details of event one from event two after the fact
21:04 Strict per incident data segregation into its own channel and document is required
Max Levchin · Feb 5, 2025
Postmortems only actually happen if there is a named directly responsible individual whose manager knows it's now part of their job, a one-to-two week deadline, and a weekly review
Without assigned responsibility and a review forum the practice degrades into a nice idea nobody executes
23:13 Named owner tight deadline and weekly review are required for postmortems to happen
Max Levchin · Feb 5, 2025
Good postmortems are matter-of-fact and focus on events and decisions, never on the individual
The event is in the past and unchangeable; the useful output is a systemic fix (e.g. adding a guard on the switch) rather than blame
24:15 Good postmortems focus matter of factly on events and decisions never the individual
Geoff Charles · Dec 6, 2023
PMs are usually enamored with their solution and fail to write down the hypothesis behind it, which is why failed products get blamed on design, marketing or sales when the real cause was a wrong hypothesis
Without a written-down hypothesis about the customer, problem and solution, you can't reflect on why the product failed or learn from it
31:30 Unwritten hypotheses explain why failed products get misattributed to execution
Geoff Charles · Dec 6, 2023
Postmortems should be blameless, cross-functional (including non-engineering teams affected), fact-driven, and end in clear timely action items
The teams that bear the consequences of a screw-up are often customer service or account managers, not engineering, so bringing them together with the data builds trust and surfaces better fixes
33:09 Blameless cross functional fact driven postmortems with clear action items
Your assistant can query this graph directly — 6 positions here, 19,646 across the corpus. Add 996.fm over MCP.