Decides fast. Irreversible.
At the line, control decisions happen in milliseconds and physically commit. If the surrounding context is not captured at that instant, it is gone — there is nothing to reconstruct afterwards.
ms · irreversibleHow regulated decisions actually happen — and where evidence has to be captured to remain reviewable later.
GMP is a time, context and decision problem — not a documentation problem.
The closer to the physical process a decision is made, the faster and more irreversible it becomes — and the harder the evidence is to reconstruct later. Switch perspective above; filter by decision layer to focus on one system level.
At the line, control decisions happen in milliseconds and physically commit. If the surrounding context is not captured at that instant, it is gone — there is nothing to reconstruct afterwards.
ms · irreversibleRelease, deviation and disposition decisions are deliberate, reviewed and signed. They are slower by design — accountability matters more than speed.
reviewed · signedCloud and analytics layers see aggregated history. They can surface a signal or trigger a review, but a GxP-relevant decision is never made far from the process.
triggers, not decisionsContext exists for milliseconds to seconds. If it is not recorded at decision time, it cannot be reconstructed — the value loses its meaning the moment it passes.
ms – sDecisions span a batch or a shift. Intervals are not abstractions — they are commitments that define meaning, responsibility and comparability.
min – hBaselines and trends are built over days to months — from structured snapshots, not raw logs. This is the reference live values are compared against.
days – monthsEvery record carries the identity behind it — which client, user or service triggered the event, and under which permissions.
Evidence is captured when the decision happens — not assembled after the fact from scattered logs.
The structured record is the source of truth, not a copy or a transformation that can drift from the original.
Values are checked against context and baselines, so a deviation is recorded as a deviation — not silently passed through.
Missing context is made explicit — deferred and queued — rather than dropped, so the record stays complete and consistent.
Hash-chained, append-only records remain readable and retrievable when an auditor asks months after the event.
Inline enforcement at the edge. The policy that allowed or blocked the message is recorded with its version.
inline · versionedProtect downstream systems from floods while keeping every throttling decision visible in the trail.
protectiveWhen required context is missing, the action is held in a reviewable queue instead of being silently skipped.
deferred · retryableRoute GxP-relevant decisions to a human gate — review, approval and sign-off become part of the evidence.
human gateSurface anomalies and recommend a review. The cloud never auto-decides a regulated action on its own.
advisory onlyAn OEE / temperature value arrives over MQTT from the filler on line 1.
Line, machine, batch and metric are resolved from the topic and payload.
The expected reference value is fetched from the historical context feed.
Live vs. baseline is compared; the deviation and its severity are computed.
The active policy decides accept, block, rate-limit or escalate — recorded with its version.
If a baseline is unavailable, the calculation is queued for review rather than skipped.
Value, context, baseline, deviation and decision are hash-chained into one reviewable record.
This is a conceptual model of how regulated decisions and evidence relate across system layers. It informs the TrailMQ design; it is not a certification claim. Read the underlying concepts →