Postmortem practice — comparison with PostHog

Status: draft scaffold. Research note. Compare our current practice against PH's, propose deltas.

What we have today:

• docs/postmortems/README.md — entry point
• docs/postmortems/_TEMPLATE.md — template
• docs/postmortems/2026-03-16_onboarding-currency-regression.md — one real PM (the only one so far)

PH's approach

TODO — fill in after a closer read of their handbook. Things to look for:

• Trigger criteria (what counts as an incident worth a PM).
• Template sections — do they have "5 whys", "contributing factors", "what went well"?
• Authoring cadence — within X days of the incident.
• Blameless framing language.
• Action items — owned by whom, tracked where, reviewed when.
• Public-by-default? They write a lot of theirs in the open.

Our current template

TODO — read _TEMPLATE.md and list the sections we have vs the PH set above.

Deltas worth adopting

TODO. Probably candidates:

• Explicit "what went well" section (we likely don't have one).
• Tagging by surface area (portal / mobile / api / infra) for pattern-spotting.
• Action item review cadence — re-read PMs quarterly to spot repeated causes.

Deltas worth rejecting

TODO. PH's volume justifies a lot of structure; we have one PM per quarter. Be careful not to import ceremony.

Action items from this comparison

TODO. Will land back into _TEMPLATE.md once decided.