Skip to content

Retrospective

60 minutes. Trio plus anyone whose work is being examined. Three questions, in order. One change — owned, dated, testable. The session that compounds the chain rather than listing it.

When

  • Within the week of the Signal reading session.
  • Calendar-locked. Held even if the cycle felt smooth — especially if it felt smooth.

Who

  • The trio — PO, Tech Lead, Designer.
  • Anyone whose work is being examined — Developer, QA, CS Lead, depending on what the signal surfaced.
  • (Optional) a peer from another team as a fresh-eyes observer for high-stakes retros.

Time-box

60 minutes. Hard cap. Five action items implemented zero is feelings; one action item implemented is compounding. The discipline is the limit.

Inputs

  • The signal reading from this cycle.
  • Any postmortems from this cycle.
  • The prior retrospective's change (so we can check if it held).

Agenda

TimeWhat
0–5 minRead the signal reading aloud. The room anchors in what reality answered.
5–25 minQuestion 1: Which chain links held? Round-robin so everyone speaks once before anyone speaks twice. Name the practice, not just the outcome. "The amigos session caught the negative-balance case" is a link holding. "Things went well" is not.
25–45 minQuestion 2: Which chain links broke? Trace to a level, not a person. Scenario-gap? Observation-mismatch? ADR drift? Prediction not checked? Each broken link names a structural gap.
45–55 minQuestion 3: What is the one change we are making? One. Owned by name. Dated. Testable. "We should communicate better" is not a change. "The amigos template gains a financial-boundary prompt by Thursday, checked at next retro" is.
55–60 minCheck the prior retro's change — did it hold? Pinned in the team channel; linked from the next cycle's kickoff.

Outputs

  • One change — named, owned, dated, testable.
  • A short retro note with: what held, what broke, the change, prior-change status. (Template.)
  • A calendar slot for Model update session — same hour as this retro, ideally.

What good looks like

The retro produces a small change — and the next cycle starts from the slightly-sharper-chain. Three to five cycles in, the team finds itself reaching for practices it didn't have a name for before. That is compounding.

The session is dry and honest. Not therapy. The conversation about feelings, if it needs to happen, happens in a separate room. The retro outputs an artefact and ends.

Anti-pattern

The retro produces a list. "Eight things to improve, none owned, all forgotten by Tuesday." Lists compound to nothing. Fix: the discipline is one change. The second thing goes to the next retro. A retro that consistently produces lists has a chain that consistently doesn't improve.

A second anti-pattern: the broken-link discussion drifts into blame. Names get used; the team gets defensive; structural fixes aren't named because the conversation became personal. Fix: the chain-aware labels (scenario-gap, observation-mismatch, adr-drift, integration-gap) exist precisely for this. Name the level. Never the person.

A third: silence in the retro is treated as agreement. Three people speak; two are quiet; the change goes through with two angles dark. Fix: round-robin every question. Silence in a retro is a system signal — what is the system not letting people say?

See also

200apps · How We Work · NWIRE