Skip to content

Model update session

30 minutes. Held the same hour as the retro. Four mechanical moves: close witnessed assumptions, add new ones, link the signal reading to the brief, sharpen open questions. The session that compounds learning — and the one most teams skip.

When

  • Immediately after the Retrospective — same hour, before the room loses momentum.
  • If skipped today, scheduled within 48 hours. Past that, the model update doesn't happen — and if the model update doesn't happen, the learning didn't happen.

Who

  • PO — primary; holds the artefacts.
  • (Optional) Tech Lead if the cycle surfaced architectural assumptions that need updating in technical artefacts.
  • (Optional) Designer if a design-system assumption shifted.

Time-box

30 minutes. Mechanical work, not creative — open the Initiative Brief and Feature Brief side-by-side and make the four moves.

Inputs

  • The signal reading.
  • The retrospective output.
  • The Initiative Brief.
  • The Feature Brief.
  • The "What we have NOT yet witnessed" sections of both briefs.

Agenda

TimeWhat
0–10 minMove 1: Close the assumptions witnessed. Open the brief's "Not witnessed" section. For each item that this cycle witnessed: mark closed, with evidence and date.
10–18 minMove 2: Add new assumptions. What did this cycle surface that we hadn't considered? Add them with status (observed / not yet observed) and implication.
18–22 minMove 3: Append the signal reading. Add a link from the brief's prediction section to the signal reading. The brief is now complete: problem, solution, whether the solution worked.
22–28 minMove 4: Sharpen open questions. Strike answered ones; rewrite refined ones; add newly surfaced ones. The set of open questions is now the next discovery's agenda.
28–30 min(If retro found a template/checklist gap) — open the template, make the change, file the PR. The retro change becomes a chain-artifact change in real time.

Outputs

  • Updated Initiative Brief — assumptions closed/added, questions sharpened — committed.
  • Updated Feature Brief — signal reading linked — committed.
  • (If applicable) template or checklist change committed.
  • A short model-update record (template) — the change log of what moved this cycle.

What good looks like

The work is small. Each move takes minutes. Together, over cycles, they are why a year-old team writes briefs a six-month-old team couldn't — because every cycle's learning is alive in the artefacts the next cycle starts from.

The test for whether the model update happened: does the next brief change because of what this cycle revealed? If yes, the update worked. If no, the update was wiki-shaped, not chain-shaped.

Anti-pattern

The session feels redundant after the retro and gets skipped. The retro produced learning; people nod; nobody writes it down where the next cycle will see it; three weeks later the next cycle begins with the same misunderstandings. Fix: the model update is mechanical, not insightful — it's the act of writing the insight down where it survives the conversation. Skipping it is skipping the only step that compounds.

A second anti-pattern: the update becomes a wiki entry, not an artefact change. A new Confluence page; a fresh Notion doc; a Slack thread with the learnings. None of these are in the path the next cycle reads. Fix: edit the brief. In place. Versioned. The next cycle reads the brief, not the wiki.

A third: no owner. Everyone assumes someone else does the update; nobody does. Fix: the PO owns it. Explicitly. Calendar-locked.

See also

200apps · How We Work · NWIRE