Observability and Diagnostics

Lab observability connects recommendation, plan, operational readiness, execution handoff, observation, and outcome. These are durable scientific and operational records. They explain more than a process log because they preserve the reasons an assay was accepted, deferred, refused, or interpreted.

Inspect the operational chain

Boundary	Diagnostic signals
Recommendation intake	candidate and evidence identifiers, review posture, unresolved questions, requested follow-up
Assay planning	assay rationale, expected information gain, sample requirements, controls, review gates
Risk and feasibility	risk findings, predicted failure risk, missing controls, feasibility and readiness scores
Capacity and scheduling	feasible and deferred batches, instrument and staff availability, slot utilization, schedule pressure
Materials	required quantities, inventory sufficiency, reservation state, material-blocked candidates
Handoff validation	accepted and rejected assay identifiers, blockers, required next actions, execution refusal
LIMS export	source-to-destination field map, record count, readiness state, and explicit loss notes
Execution outcome	assay observations, QC feedback, batch assessment, failure triage, and follow-up directive

Use candidate, plan, batch, assay, sample, and observation identifiers as the join keys. Every scheduled assay should trace to a plan and recommendation; every observation should trace to an assay; every follow-up should trace to an outcome. An orphaned identifier is an integrity issue, even when its individual record validates.

Separate blockers from failures

A deferred batch caused by capacity or materials did not fail experimentally. A responsible handoff refusal is a safety outcome, not a service outage. A LIMS loss note identifies an interchange limitation. A completed assay with adverse QC is an observed scientific or operational failure. Preserve these states so throughput metrics cannot convert blocked or refused work into unsuccessful execution—or hide it as success.

An incident record should include the upstream decision and evidence references, plan and batch identifiers, lifecycle state, risk findings, readiness and practicality, missing controls or resources, schedule pressure, handoff blockers, export loss notes, observations, and outcome classification.

Diagnostics are complete when a reviewer can explain why work was or was not scheduled, what was transferred to the execution system, what actually happened, and which evidence authorizes the next action.