ISO 8000-120 addresses provenance for characteristic master data. Its published scope excludes configuration management and change control, and it does not supply a general telemetry-lineage contract. GFT therefore uses it to inform provenance discipline within that stated boundary. The analytical event lineage recommended below is a separate design proposal, not a claim of compliance with a standard whose scope is narrower.[S41]
GFT recommends recording the source observation reference, acquisition configuration revision, applicable mapping revision and transformation version whenever a value changes meaning. A converted temperature, a counter difference and a classified defect have different transformation dependencies. The lineage should make those dependencies inspectable without requiring the reader to infer them from a final dashboard label. Unchanged raw evidence remains the anchor for reviewing subsequent interpretations.
An illustrative counter calculation may start at 1,200 and finish at 1,248, producing a nominal difference of 48 counts. That arithmetic is meaningful only under the assumed conditions of a continuing counter instance, unchanged count definition and no intervening reset. If a reset is recorded, the simple subtraction no longer establishes output. GFT would retain the observations, mark the interval unresolved and apply a separately approved reconstruction rule if independent evidence permits one.
Reconstruction should declare its authority and uncertainty. GFT recommends distinguishing an observed count, an arithmetic result from accepted observations, and an estimate reconstructed from another record. A manual correction should identify the role, reason, affected period and superseded interpretation. It should not erase the source values or silently turn estimated production into measured production. The display can remain concise while exposing those distinctions through its evidence view.
For an illustrative defect event, the lineage may include the image reference, acquisition setup, model version, decision threshold and human review outcome. This does not require keeping every possible artefact indefinitely. It requires a retention decision matched to the use and a clear explanation when evidence expires. Storage architecture and retention implementation belong to the later historian chapter; this chapter defines what relationships that implementation must preserve.
GFT would express event quality through separate dimensions where the decision requires them: measurement validity, identity confidence, clock state, process-context status and delivery freshness. A reading may have valid measurement quality while its order assignment remains unresolved. Another may have complete identity but arrive too late for immediate intervention. Combining all those states into an unexplained good flag would conceal which use remains defensible.
The proposed acceptance exercise should follow a representative event from physical capture through an intentional transformation and into a decision. The reviewer should recover the original observation, identify the applicable asset assignment and explain every timestamp. The exercise should then repeat with a gateway restart, an assignment correction, a delayed replay and an invalid clock state. These are recommended tests, not reports of completed GFT operations.
For each test, the expected interpretation should be approved before execution. Evidence of correct rejection matters alongside evidence of correct acceptance. An unknown identity should remain unresolved. An ambiguous time association should remain ambiguous. A corrected mapping should produce an explainable revision. The factory should be able to reproduce both the original decision view and its later correction without claiming that late knowledge existed at the time.
GFT recommends making release authority part of the handover. The metric owner should accept the definition, the acquisition owner should accept capture and clock evidence, and the receiving application owner should accept interpretation and exception behaviour. Conditional acceptance should state which use is permitted and which evidence remains missing. A signal may be suitable for a maintenance trend while remaining unsuitable for product allocation. Recording that restricted permission gives the factory a useful intermediate state without presenting partial qualification as universal readiness.
The resulting handover is a compact contract: an identity dictionary, effective-dated mappings, declared timestamp semantics, uncertainty rules, event-promotion criteria and retained transformation references. It gives the following capture chapter a clear obligation: acquire observations without losing these distinctions. It also gives later storage and dashboard designs an acceptance test. A displayed value must be traceable to the production evidence that permits its interpretation.