Topic 1.2 — Metrics and queries · cram sheet

Flow metrics — the definitions

Diagnosis: short cycle time + long lead time = the team is fast but overloaded. Limit WIP; don't just hire.

Planning

• Velocity is a planning input, never a target. Targeted → estimates inflate, prediction dies.
• Widening band on a cumulative flow diagram = a growing queue in that state.
• Query types: flat list · work items and direct links · tree of work items (hierarchy).

Development

• Time to first review is usually the largest hidden delay — larger than build time.
• Misleading: lines of code, commits per developer, individual velocity. All punish helping colleagues.
• High churn = unsettled design in that area.

Testing

• Flaky test rate belongs next to pass rate. 99% pass + 4% flaky ≠ healthy — every failure becomes ambiguous.
• Gate on coverage of changed lines, not repo-wide.
• Coverage shows lines executed, never whether anything was asserted.

Security

• Age of findings beats count. 50 findings under a week ≫ 5 findings two years old.
• Split MTTR by severity — aggregation hides triage failure.
• Dependency currency is a leading indicator of future remediation speed.

Delivery & operations

• Deployment frequency alone is gameable — pair with change failure rate.
• Smaller batches reduce per-deployment risk.
• Alert on user-facing signals; diagnose with infrastructure.
• Report p95/p99, not averages. Azure Monitor logs use KQL.