Appendix D - 1. The Operator's Dashboard

James C. Davis

Appendix D - 1. The Operator's Dashboard

The dashboard collects the primary metrics used to operate a Governed Engineering Environment. Formative metrics steer work while it is in progress. Summative metrics certify the finished system. Some metrics serve both purposes.

Table D.1-1 gives each metric in two bands — the formative metrics above the divider, the summative verdicts below — with what each one counts, when to read it, and its healthy direction. Scan the band you need. The 2.5 reference is the companion one level down: read it to steer a single loop iteration, read this to steer or grade the whole program.

Table D.1-1. The Operator's Dashboard — every metric the build acts on, in two mode bands: the formative metrics you steer by while the work is in flight, then the summative verdicts you certify the result with at maturity.
Metric	Mode	What it counts	When to watch	Healthy direction	Defined in
Formative — measured during the work, to steer the next step
Missing-Model Metric	formative	Fraction of tests whose exercised code traces to no model claim — the unmodelled surface — plus its drain curve.	After each model-loop Epic; steer the next at the biggest orphan cluster.	Drains toward 10%-or-under (56% to 7.89% over nine re-runs).	2.5
Velocity	formative	Commits per week.	Watch the dip where velocity buys hardening.	Roughly linear; a hardening dip is expected, not alarming.	5.2
Churn	formative	Lines added and deleted per week per path.	Reads which build phase you are in.	Peaks at mechanization, then collapses as the environment stabilizes.	5.2
Model-sync efficacy	formative	Whether the drift and parity gates keep model equal to code.	Watch that map-equals-territory holds.	Gates stay green.	3.8
Grammar coverage	formative	Whether the generator exercised every production of the input grammar.	Watch for corpus holes no line-coverage number reveals.	Rises toward full grammar exercise.	4.6
Model-claim coverage	formative	Whether generated inputs drove every declared invariant, transition, and edge.	The saturation oracle for generative validation.	Rises toward full claim exercise.	4.6
Summative — measured at maturity, a verdict on the result
MBSE navigation token-savings	summative	Tokens spent to reach an answer, model on versus off.	Certify the model earned its context budget.	Lower with the model on.	3.1
Support ratio	both	Support-apparatus LoC (tests, lints, docs, infra, tooling) over production LoC.	Watch the apparatus keeps leading as feature work resumes.	Leads production; settles around 3x it at maturity.	5.2
Control growth	both	Project-specific lint files and gate scripts, per window.	Watch the environment still accrete controls.	Climbs steadily (lints 0 to 747; gates 0 to 102).	2.3
Epic-closure rate	both	Epics moved into the closed set per week — the finishing rate, not the commit rate.	Whether an operating-mode shift converts to durable throughput rather than raw output.	Rises when the environment absorbs autonomous loops; flat velocity beside a rising closure rate is the healthy anti-decay shape.	5.2

Each metric's mode and the rationale for its formative-or-summative call live in the dashboard's model file, which projects this table and holds the page equal to it.