DocumentationBuild pipeline health check

Build Pipeline Health Check Scoring

Reference contract for scoring and interpreting the build pipeline health check.

Last reviewed
Version
1.0
  • assessment
  • build infrastructure
  • planning

This reference defines the scoring contract for the build pipeline health check. It makes a current delivery path easier to discuss and sequence; it does not certify a pipeline or replace engineering judgment.

Required dimensions#

Answer all six dimensions, once each, in the table order shown. Each answer is an integer score from 0 through 4. The order also resolves action-priority ties.

OrderDimensionAssessment focus
1Source and input controlWhether eligible source, dependencies, and build inputs are explicit and controlled.
2Build definition and environmentWhether targets, tooling, worker setup, and replacement behavior are declared.
3Artifact identity and retentionWhether outputs are immutable, traceable, retrievable, and accompanied by evidence.
4Compatibility and promotionWhether compatible candidates are selected and promotion has enforceable gates.
5Recovery and rollbackWhether common failure paths and a prior proven release can be restored.
6Ownership and observabilityWhether failures are visible, assigned, and actionable without hidden personal knowledge.

No result/export is available until all six dimensions are answered.

Shared score choices#

Use the same choices for every required dimension. Select the description that best represents normal operating behavior, not an exceptional success or failure.

ScoreChoice
0Unknown or ad hoc
1Manual and person-dependent
2Partly documented
3Repeatable and recoverable
4Proven and observable

A dimension may be well understood and still score low when its outcome depends on an undocumented command, a specific machine, or a particular person. Conversely, documentation alone does not justify the highest score unless the path is routinely proven, observable, and tested.

Calculation and bands#

Add the six selected scores, then calculate the percentage as round(sum/24*100). The calculation rounds once after dividing the total by 24. Do not round individual dimensions or substitute a weighted score.

PercentageBand
0–39Stabilize
40–69Controlled/recoverable
70–84Repeatability/ownership
85–100Maintain/test edges

The displayed band is determined from the rounded percentage. It is not determined from the average score text or from a subjective interpretation of the assessment.

Action priority#

The lowest two dimensions determine actions. Ties use table order. For example, if several dimensions have the same lowest score, select the earlier two in the required-dimensions table before considering later rows.

Treat those two dimensions as the starting boundary for planning. Define a small observable improvement, an owner, acceptance evidence, and a recovery check for each. Do not use a stronger overall result to defer an unowned or unrecoverable weakness in one of the selected dimensions.

Boundary test vectors#

These vectors verify the calculation and band transitions. Scores appear in required-dimensions table order.

ScoresSumResultBand
0, 0, 0, 0, 0, 000Stabilize
1, 1, 1, 1, 1, 4938Stabilize
1, 1, 2, 2, 2, 21042Controlled/recoverable
2, 2, 3, 3, 3, 31667Controlled/recoverable
2, 3, 3, 3, 3, 31771Repeatability/ownership
3, 3, 3, 3, 4, 42083Repeatability/ownership
3, 3, 3, 4, 4, 42188Maintain/test edges
4, 4, 4, 4, 4, 424100Maintain/test edges

Interpretation limits#

The score is a planning signal only. It helps identify the next delivery boundary to strengthen and makes the rationale visible. It does not predict release outcomes, guarantee readiness, replace validation evidence, or rank teams. Reassess after a meaningful change or a recovery exercise so the recorded state continues to describe normal operating behavior.