Metrics Tracking

Step 2: Collect the Raw Data

For each metric, identify the data source:

• Sprint tracking tool (velocity, commitment)
• Test management tool (coverage, defect counts)
• Monitoring dashboard (latency, error rates, drift)
• User analytics (adoption, feature utilization)
• Lifecycle artefacts (artefact completion, gate outcomes)

Collect data at the defined cadence (see metrics catalog for each metric's collection cadence).

Step 3: Calculate the Metric

Use the formula from references/metrics-catalog.md. Do not estimate or approximate — use actual measurements. If data is unavailable, document the data gap as a risk and note the metric as "not available — data gap."

Step 4: Compare Against Thresholds

For each metric:

• Compare the calculated value against the defined target threshold
• Classify as: On Target / Warning (within 10% of threshold) / Breached (below threshold)
• Any breached metric triggers an action: investigate root cause, define corrective action, assign owner and due date

Step 5: Produce the Metrics Report

Generate the metrics report using the appropriate template:

• Sprint metrics: include in sprint review record
• Phase metrics: include in gate pack
• Operations metrics: use templates/phase-6/service-report.md.template
• AI metrics: use templates/phase-6/ai-monitoring-report.md.template

The report must include: metric name, value, target, status (On Target/Warning/Breached), trend (improving/stable/degrading), and action (if any).

Step 6: Escalate Breached Metrics

If any metric is Breached:

• Create or update the relevant risk register entry
• Notify the metric owner
• Define a corrective action with timeline
• Track corrective action in the improvement backlog
• At next sprint review, re-evaluate the metric

Step 7: Record in Evidence Index

After producing the metrics report:

• Register it in the evidence index with the gate it satisfies
• Note which metrics were not available and why

Key Principles

1. Metrics without targets are vanity — always define target thresholds before collecting.
2. Trends matter more than snapshots — a metric at 80% of target but improving is better than one at 95% and degrading.
3. AI metrics require baselines — establish model accuracy and latency baselines at Phase 4 Gate D; deviations in Phase 6 are measured against these baselines.
4. Governance metrics catch process failures — a high waiver rate signals the team is under-investing in readiness.
5. Do not interpolate missing data — a data gap is a risk to be documented, not a number to be guessed.

Reference Materials

• references/metrics-catalog.md — Complete catalog with formulas, targets, collection methods, and owners
• templates/phase-6/service-report.md.template — Operational metrics report template
• templates/phase-6/ai-monitoring-report.md.template — AI monitoring report template

Quality Checks

• All reported metrics include formula, actual value, target, and status
• Breached metrics have an assigned corrective action and owner
• AI metrics include baselines established at Gate D
• Metrics report is registered in the evidence index
• Collection cadence is being met (no gaps of more than 2 sprints for delivery metrics)