Machine run observations

Machine run facts describe individual observations attached to a production snapshot. The team can retain sequence, description, quantity, unit, observation date, production date and source reference alongside the outcome. Session, conversion, machine and exposure figures can also be recorded when the measurement requires them. The linked parent keeps the fact connected to the production source being checked.

A production analyst can read facts marked Pending before invoking validation. Accepted and Rejected are separate outcomes, so a rejected fact remains explainable without silently disappearing from the record. A source reference gives the analyst an identifier to reconcile with the upstream observation when an amount or quantity is questioned. The fact record does not substitute for the source capture; it identifies the observation used in this working cycle.

• Sequence. Keep observations in a deliberate order so reviewers can inspect a repeated or corrected submission.
• Outcome. Pending, Accepted and Rejected make the treatment of each fact visible to the team.
• Parent. Every fact belongs to the production snapshot whose revision is being reviewed.

Define which upstream events count as a machine run observation and how their source references are assigned. Agree the units used for quantities and amounts before importing a mixed feed. The Snapshot Owner and Production Analyst should decide who reviews a Pending outcome and how a rejected observation is explained. If consent basis matters for the underlying session data, agree its source and who may inspect it. Keep that field meaningful to the team rather than filling it with a generic label.

Suppose a source has observations from a shift and one event was repeated by the upstream collector. Open the parent snapshot and compare the facts by sequence, observation date and source reference. Mark the fact’s outcome according to the reviewed evidence. Invoke validation for the intended source revision, then inspect the operation run and its numbered steps. If validation is retried with the same operation key, the team should find one explainable run outcome rather than a second unexplained result. Ask the owner to show why the repeated event was Accepted or Rejected and which observation date belongs in this production window. Keep the original source reference available when the analyst checks the next capture.

• Machine facts needing review — Pending observations with their sequence, quantity, date and parent source.
• Production snapshot queue — The source records to which the observations belong.

Start with observation validation, then check source coverage. Use the implementation plan to agree how upstream references enter the system.