Skip to main content

Metrics framework

The measurement standard. One framework, feeding one platform with many views (not eleven dashboards). Baselined during StudyBuddy so "velocity compounds" is measured, not asserted.

What we measure​

  • DORA four keys — deployment frequency, lead time for change, change-failure rate, MTTR. Emitted by the pipeline itself.
  • SPACE dimensions — satisfaction, performance, activity, communication, efficiency — the human/flow side, so we don't optimise throughput into burnout.
  • AI-specific measures — the ones unique to building this way:
    • Agent first-pass acceptance rate — how often agent output is accepted without correction.
    • Rework rate on agent PRs.
    • First-time gate pass rate.
    • Token cost per shipped change — the unit economics.
    • Capital-deposit rate — runbooks, prompts, playbooks and gotchas added per loop. This is how we quantify compounding.

The rules​

  • Metrics inform, never rank individuals. This anti-gaming rule is baked in — the moment a metric is used to rank people, it stops measuring reality and starts measuring fear.
  • Auto-fed, not hand-reported — from git, CI/CD, the registers and agent telemetry. No manual reporting overhead.
  • One platform, many views — STOP progress, savings-vs-AI-spend, DORA, quality, security, AI cost, adoption, commercial proof-points — all from the same source, filtered per audience.

Standards referenced: DORA, SPACE, DX Core 4. The platform build itself is a later pass (TASKS.md §6); this standard defines what it must compute.