Illustrative rating examples

These examples clarify the distinction between states 1 through 4; they are discussion aids, not validated thresholds or extra requirements. They are written against the current assessment items and have not yet been checked in pilot sessions. Rate the evidence for the practice in the project context: team size, architecture, risk, and operating needs can make different methods equally effective. A team does not need a particular tool or ceremony to earn a state.

Code review (item 4.13)

State	Example evidence
1 — Ad-hoc	Some changes receive review when a teammate is available, but expectations and coverage vary. A recent change may have merged without another person's review.
2 — Defined	The team has an agreed review approach and uses it for the changes it identifies as needing review. Reviews commonly leave a record, such as an approval or discussion in the change history.
3 — Managed	The team can show that its review approach is followed over time, checks exceptions, and addresses recurring delays or missed reviews.
4 — Optimizing	The team examines review outcomes and friction, makes changes based on evidence, and checks whether those changes improve quality or flow without undermining review effectiveness.

Recovery plan (item 5.10)

State	Example evidence
1 — Ad-hoc	Recovery depends on individual knowledge or improvised steps; people cannot reliably describe how to restore the service.
2 — Defined	Relevant recovery steps and responsibilities are recorded, and the team can identify where the needed backups or recovery inputs come from.
3 — Managed	The team has practiced recovery at a cadence appropriate to its risk, records results, and follows up on failures or gaps.
4 — Optimizing	The team uses recovery results and operational changes to improve the plan, then verifies that the changes reduce recovery risk or effort.

Applying the examples

• Choose the highest state supported by the evidence discussed. A written process alone does not demonstrate that it is followed; a successful one-off does not demonstrate consistency.
• Treat the examples as anchors for discussion, not a checklist. Equivalent evidence can support the same conclusion, and the amount of evidence should fit the risk and context.
• Record uncertainty or disagreement instead of forcing a rating. Use the pilot method to compare independent ratings of shared evidence and record where these examples need clarification.
• These examples do not establish inter-rater reliability or validate the score. Update them only after pilot findings are recorded; semantic changes follow the versioning policy in the changelog.