How is the Living Compliance Index calculated?
The Living Compliance Index (LCI) is a 0-to-100 measure that does not score what happens inside a facility, but when its records are born: did the evidence arise the moment the work was done, or was it manufactured in the preparation week before the audit? The index answers four questions about every record — Was it found? When was it born? With its evidence? And where? — not the quality of the work, nor its speed.
The reference methodology of the goiso platform — version v0, July 2026.
This methodology is published for review and to establish precedent. The weights and constants shown below are starting values under field calibration, and may change across later versions — each documented with its date.
The governing principle: we measure birth, not quality
No raw count enters the index. Every quantity crosses into it converted to a ratio, a lag, a match, or a trajectory. A facility is neither rewarded for having few incidents nor punished for having many; a bad day well managed raises the index rather than lowering it. The membership test is decisive: an element enters the index only if its absence would mean a missing piece of evidence on audit day had the preparation week been cancelled.
Results-blindness — three rules
The index is built on a deliberate blindness to results, through three rules that keep it from rewarding luck or punishing honesty:
1Counts are never scored
The number of incidents or failures never enters the score. An event merely opens an opportunity to prove management; what is measured is how it was handled, not that it occurred.
2A result differs from a discovery
Result-events (incidents, breakdowns) are matters of chance, and a zero over ninety days can be honest. But discovery-items (inspection findings, observed hazards, corrective actions) do not befall — they are seen; they are the output of the system's eye. Their complete zero is a closed eye or a theatrical inspection, not genuine calm.
3Existence guides diagnosis, not the score
The presence or absence of a thing only distinguishes a documentation gap from honest calm — it does not directly raise or lower a score. This is the work of the double-zero gate below.
The four-dates model
Every record and its evidence carry four times, from which two gaps are derived that reveal the honesty of the documentation:
Record opened
The moment the card is created in the system.
Closed
The moment the loop closes (the work done and delivered).
Evidence uploaded
The system stamp that cannot be edited — when the file actually entered.
The witnessed-event date
The day the document itself records (the meeting held, the inspection performed, the reading taken), extracted from its text, not from its upload stamp.
The system gap (t1 minus the declared time of occurrence) measures how late documentation trails the event. The birth gap (t3 minus t4) is the only detector of the “migrating parallel system” pattern: a facility that worked on paper for months, then transferred it in one batch with an instant upload — its system gap is a deceptive zero, and the document's true date exposes it.
A strict rule: a document's date is accepted only if it appears explicitly and unambiguously, with a verbatim quoted passage from the document; otherwise it is neutrally excluded. Precision is favoured over recall, and extraction dates are never edited by a human hand — for the measured party does not own its own ruler.
The four pillars and their weights
The index is composed from four pillars by a deliberately weighted geometric mean: a dead pillar sinks the whole, for life requires every organ. The weights below are v0 starting values.
Birth
The synchrony of documentation with the work: how late the evidence was uploaded after its event date, how late the incident was recorded after it occurred, and whether evidence production spreads across the days or piles into a single week.
Existence
Whether loops close (every incident yields a structural response), and whether the scheduled is done on time and documented, even if honestly late. It holds the double-zero gate.
Evidence at the source
The share of closures that carried their evidence at the moment of closing, not weeks later — the proof is born with the closure, not appended to it.
Coverage
The share of locations actually touched by an operational record, against the declared locations — it prevents polishing one corner and neglecting the rest.
The double-zero gate
When a measurement window holds zero discovery-items and zero result-events together, this is not calm but a documentation gap: an eye that sees nothing and at which nothing occurs is a closed eye. The existence pillar then drops to a calibrated floor. The rule: “two zeros never meet in reality — they meet only on paper.”
The integrity guard
A multiplier from 1.0 down to a floor of 0.8, surfaced as flags rather than a hidden score. It catches synchronized forgery: batches of evidence sharing the same second across different cards, retroactive edits on closed records, impossible dates, evidence appended after delivery. “Living-documentation theatre” is synchronized forgery, not living compliance.
The four bands
The daily score is smoothed by an exponential moving average with a half-life of about thirty days — the index moves like a pulse, not a counter, so a week of illness does not disgrace the warrior. The smoothed score then falls into one of four bands:
Why “preliminary”?
The index runs today in a silent shadow phase: computed from existing operational data without being shown to anyone, so that we gather its real distribution from actual facilities and calibrate the weights and thresholds from it. The philosophy, the structure, and the vocabulary (results-blindness, the four dates, the bands) are stable; only the numbers are under calibration. We publish the methodology now, dated, to establish its precedent and open it to review — not because its calibration is complete.
This index measures the concept of Living Compliance — read its reference definition.
Does the index punish a facility for having many incidents?
No. The number of incidents never enters the score — that is the results-blindness principle. The index measures how an event was managed and when it was documented, not that it occurred; a facility that records its incidents honestly and closes its loops with timely evidence ranks above one whose records are empty because its eye is closed.
What is the difference between the system gap and the birth gap?
The system gap measures how late the record was created in the platform after the event, and is available from system stamps. The birth gap compares the evidence's upload stamp with the event date written inside the document itself, and it alone exposes anyone who worked on paper for months then uploaded everything in one batch — whose system gap is a deceptive zero while the document's date betrays them.
Why a geometric mean rather than an arithmetic one?
Because living compliance requires all its pillars together; a dead pillar must sink the index, not be offset by another's excess. The weighted geometric mean does exactly this: a zero in any essential pillar plunges the whole, just as a single living organ is not enough for life.
Are the weights (35/25/20/20) final?
No. They are the v0 starting values, calibrated from the real distribution of actual facilities' data in the shadow phase. The structure and vocabulary are stable, and the numbers are documented with each version by its date, as management-system standards do.
Is the index shown to the tenant now?
Not yet. The index runs in a silent shadow phase in which it is computed without being displayed, until its calibration matures from sufficient field data. Publishing the methodology here is to establish precedent and open the door to review, not to announce a launch of the surface.