Decide what would make you stop the experiment
https://richarddurnall.com/guides/a-retrospective-that-produces-one-test/

A review date is useful, but the team also needs to know what evidence would justify continuing, changing or stopping a trial. Write that decision before the result is known. Otherwise the group can keep redefining success to protect a favourite idea.

A filled experiment card

FieldFictional team proposal

ObservationRecent requests reached testing with missing acceptance examples.

ChangeRequester and tester review the next five suitable requests before implementation.

RecordMissing details found; review time; urgent work delayed; gaps found in testing.

Stop or adaptPause if the review blocks urgent work; adapt if details remain unclear after review.

Decision pointAfter five suitable requests, review every record, including skipped and difficult cases.



Keep the review open to disagreement

Define “suitable” before selecting the requests. Record exclusions and their reasons.

Name an owner who can arrange the review and affected people who can raise the cost of the change.

Keep observations and explanations in separate columns. Fewer recorded gaps may reflect different work, not improvement.

At review, choose continue, adapt or stop and state the evidence for the choice.

If the trial was not run, record that fact and discuss feasibility. Do not invent an outcome to close the action.



Five requests is a bounded example, not a statistically justified sample size. A stop condition is a team decision for this trial, not a universal Agile rule. Regular reflection and adjustment appear in the Agile Manifesto; the experiment card and its examples are independent editorial additions, not reported results or new work attributed to Richard Durnall.



Text equivalent: Run the agreed trial → Record cost + gaps → Continue → Adapt → Stop

Observed event: __________________________

Working explanation: __________________________

Change to try: __________________________

Suitable cases and exclusions: __________________________

Owner and affected people: __________________________

Evidence including burden: __________________________

Stop/adapt condition: __________________________

Review point: __________________________

What was actually tried: __________________________

Continue/adapt/stop and reason: __________________________

Original reversible-experiment map. Branches are possible decisions, not tested outcomes.
https://agilemanifesto.org/principles.html