What you license · 04
Metrics and dashboards
Governed by instruments, not opinions. The Metrics Registry is the single definition of every measure the organization reads, the catalog states the property arithmetic behind each one, and dashboards are faces over one instrument, so two people on different dashboards are never looking at two versions of the truth, only two selections from it.
Every metric is only as true as the properties beneath it. A cycle time computed from an unstamped date or a verdict mix over outcomes nobody armed is not a weaker signal; it is a wrong one, and wrong signals cost more than none because people act on them. That is why the Delivery OS stamps every timestamp by automation, why the facilitation skills land judgments as properties at the moment they are made, and why every complication carries an honesty caption with its thresholds quoted from the register and their provenance stated.
Choose your face
One instrument underneath, a face for each role and each moment on top.
Dashboards in the Delivery OS work the way watch faces work: the instrument always knows the time, the pace, and the pulse, and the face you choose decides which of those you see at a glance and how they are arranged. A dashboard is a configured page of live filtered views over the Delivery OS databases, with skill-generated reads layered in for what views cannot compute. Every presentation of every metric is a complication, built once in the Complication Library to its registry row’s specification. When no face fits, you make your own in minutes: duplicate the face template and copy the complications you care about from the library. Copying a complication copies its configuration, never its data.
The short read
Each role acts on a handful of metrics. The registry says which.
Executives
Trends, not points: outcome verdict mix, the initiative outcome roll-up, quarterly commitment reliability, increment throughput, full-thread cycle time, the reviewed share of Done, value realization lag, the 90-day-win test, the portfolio WIP trend, and where effort is actually going against the year’s goals.
Leadership
The tier that runs delivery day to day: initiative health mix and whether each initiative’s own signal agrees with its increments, feature age-in-state, the Ready for Dev queue, blocked features and increments with owner and target resolution, the capacity worksheet, and maintenance consumption against allocation. Nothing on it ranks teams or people.
Product and Engineering Managers
The health of their pipeline: discovery aging by complexity class, intake queue depth, the two halves of cycle time, plan tracking, feature flow, outcome arming coverage, overdue outcome reviews, and whether ordering rests on fresh judgments or stale ones.
Delivery Management
Exception-shaped: what is blocked and for how long, mirror drift between Notion and Linear, overdue features, flow debt and WIP, the maintenance run-rate, RAID aging, and whether every red or amber increment carries the RAID item that explains it.
Business roles
Scoped, never softened: a business line’s slice of the same truth everyone else lives by, laddered into sentences. What is in flight for my line, whether the outcomes promised to it are being met and confirmed, whether the work it depends on is tracking, its dependency exposure, and its delivery pace.
The events
The event faces arrange the instrument around a moment rather than a role: the metrics a planning event needs in the room, the ones monthly refinement orders on, and the ones the Watch reads every morning.
The catalog
Roughly sixty metrics across nine domains, each with its definition, formula, sources, aliases, computation path, and accuracy contract.
- A
Queue and aging
Backlog, discovery, validation, and ready-to-implement age; intake queue depth; re-verification currency; the Ready for Dev queue.
- B
Flow and time
Definition-half, delivery-half, and full-thread cycle time; increment throughput; flow debt; blocked duration; feature age-in-state and blocked-since.
- C
Ordering and value
Score vintage, staleness under a superseded Definition of Value, the Cost of Delay profile mix and its evidence currency, ordering integrity, must-split sentinels.
- D
Commitment reliability
Quarterly commitment reliability, plan tracking, hard-deadline exposure, carryover split, feature overdue, target-date integrity.
- E
Outcomes and value realization
Verdict mix, arming coverage, unarmed-on-Done escapes, overdue reviews, value realization lag, reviewed share of Done, the 90-day-win test, the initiative roll-up, feature-to-outcome traceability.
- F
Capacity and predictability
Row loading against the 80% line, stated versus delivered, unplanned arrival rate, maintenance consumption against allocation, confidence calibration, plan drift, sizing accuracy.
- G
Risk
Open RAID by severity and type, RAID aging, dependency category mix, blocked increment count, health-linkage discipline.
- H
Portfolio composition and instrument integrity
WIP by state, type mix, complexity-class mix, product-line cut, goal-code investment cut, mirror drift, record completeness, initiative health mix.
- I
Code flow
Where finished code waits, first-review latency, review-load concentration, read live from GitHub and laddered into sentences for business audiences.
The catalog is built in both directions at once. Bottom-up, it walks every property in the Delivery OS and asks what it can measure. Top-down, it confirms that every metric the flow already names, in the Threshold Register, the monitor’s check catalog, the planning simulation, and the value realization practice, has its property arithmetic stated. Where the schema could compute something nothing consumed, the gallery gave it a consumer.
The threshold register
Thresholds are ratified, dated, and quoted with their provenance.
The exception-based watch only works if “aging” and “too deep” mean something agreed. The Threshold Register holds the norms every check evaluates against, by complexity class where the norm differs: discovery age, backlog age, intake queue depth against the buffer, portfolio WIP, blocked duration, plan tracking tolerance, value realization lag. Thresholds are seeded as proposals the way the methodology prescribes, ratified under Lean Governance, and revised in the portfolio retrospective, so a threshold is always a decision someone made on a date, never a default nobody remembers choosing.
The same honesty rule reaches the roadmap: committed, planned, and projected are encoded differently in every rendering, and nothing projected is ever drawn with committed-grade solidity.
See it running
Two faces over the BigBank instrument.
The Executive face and the Delivery Management face
The same quarter read at two altitudes: the outcome verdict mix and commitment reliability for leadership, the blocked list and mirror drift for the people unblocking it.
What the license contains
The registry, the catalog, the library, the faces, and the register.
The Metrics Registry schema with its faces and complication-specification properties, the Metrics Catalog as the arbiter of definitions, the Complication Library build sequences, the five role faces and three event faces with the face template, the Threshold Register with its seeded proposals, and the data-health first stamp that keeps the instrument honest from the first week. Numbers questions asked in Notion AI are answered from the registry and point to the dashboard where the live read lives. How the tiers are shaped.