Documentation / Calibration / The gage register

I — What the register is actually for

A register identifies instances rather than models

Every item used to determine whether product conforms has to be uniquely identified: the individual instrument rather than its model. The register exists so that a question asked years later, about a measurement nobody remembers taking, still has an answer.

Reference · No product claim on this page

“Mitutoyo 293-340” and “MIC-0147” are different kinds of fact. The first is a model, shared by every one of them in the building. The second is an instance, and an impact analysis can only ever be run on instances. A register keyed by model looks complete and answers nothing.

II — The ID scheme

Three rules do most of the work, and the third is the one that gets broken.

Rule	Why
Identify the instance	Impact analysis, recall and any history at all are per-instrument facts.
Mark it durably, and machine-readably	A data-matrix code and a reader at the station turn “who measured this?” from an interview into a database fact.
Never reuse an identifier	A retired MIC-0147 whose number is issued to a new micrometer silently merges two instruments’ histories, and every analysis over that range is then wrong in a way nothing detects.

Some things need their own IDs even though they are not instruments. A CMM stylus, a setting standard, a setting ring (anything that can independently cause an error) needs tracking separately, because when it turns out to be the cause, the recall follows it and not the gage it was fitted to.

III — Status, and where it should live

Calibration status is conventionally carried on the instrument: a label with the ID, the dates and any limitation, plus a void-on-removal seal where tampering matters. One label type is worth more than it looks: the limited-use label, for an instrument that passes over part of its range and not the rest. It is more honest than the two alternatives, which are scrapping a serviceable gage or passing one that should not be trusted at the top of its range.

For a connected gage there is a strictly better option: enforce status at the point of use. A sticker requires a human to look at it, and operators do not. A system that knows which instrument is sending can refuse the reading instead of relying on someone reading a date.

IV — Check-out is a metrological record

Issue and return get treated as administration, a signature so the crib knows where things are. They are worth considerably more than that, because they bound which measurements an instrument could possibly have made.

If MIC-0147 sat in the crib from March to June, then nothing produced in those months was measured with it. When that gage later fails calibration, that single fact removes three months from the exposure window, and the window is what decides how much product is in question. Very few organizations exploit a record they are already keeping.

issued   2026-03-02  → station 4
returned 2026-03-06  → crib
issued   2026-06-11  → station 4
                          ↓
found out of tolerance 2026-07-30
window = 06-11 to 07-30, not 03-02 to 07-30

V — The two failure modes nobody fixes

Personal gages. Instruments owned by the machinist who uses them sit outside the system entirely, are never calibrated by it, and are used to make real decisions about real product. The register holds no record of their existence. This is the most common and least discussed hole in a gage program.

Gages issued years ago that never came back. They keep generating due dates, which get acknowledged and rescheduled by someone who has never seen the instrument. The register reports a healthy program while describing equipment nobody can produce.

VI — What a calibration record has to carry

A certificate that says only PASS has thrown away everything that makes the record useful later. The fields below are what an interval analysis, an impact analysis and an auditor each need, and none can be reconstructed afterwards.

Field	Note
Gage instance ID	Instance, never type.
Calibration date and time	Distinct from the certificate’s issue date — the interval runs from the calibration.
Due date, and the interval that produced it	The interval is the reasoning; the date alone is not.
Performed by	Person and organization.
Procedure used, with revision	A method without its revision is not repeatable.
Reference standards used, by instance ID	With their own certificates and their due dates as at the time of use — this is what makes a cascading recall possible.
Environmental conditions	Flagged as measured, assumed, or template. The three are not the same claim.
As-found values, per check point	Numeric and signed, before any adjustment.
Acceptance limits used, per check point	So that “out of tolerance” has a definition as of that date.
As-found conformity result	In tolerance, out of tolerance, or limited.
Adjustment or repair performed	What was actually done.
As-left values and result	Numeric and signed, after adjustment.
Uncertainty per check point	A traceability claim without an uncertainty is not one.

VII — Sources and boundaries

The register described here is the practice the calibration and metrology literature converges on. Where a specific number or designation could not be verified against a primary source, it does not appear.
No clause numbers from IATF 16949 or AS9100 are cited. They are paywalled and could not be confirmed after repeated attempts, and a clause number quoted wrongly is worse than none.
SVEND does not sell gage crib or calibration management software. This page exists because the questions it answers decide whether a measurement record is worth anything.

When a gage fails calibration → Calibration intervals → What a Digimatic frame carries →