Error Model

Lab represents planning blockers, operational failures, scientific outcomes, and promotion conflicts separately because each demands a different response.

Condition	Representation	Response
Invalid design	Experiment design or plan validation issue	Correct cohorts, controls, assays, dependencies, or metadata
Dependency cycle	Dependency cycle or integrity report	Repair the plan before scheduling
Resource blocker	Readiness report naming instruments, materials, staffing, cost, or backlog	Defer or re-plan; do not mark execution ready
Missing control or provenance	Readiness blocker or artifact contract issue	Restore the required control or lineage before handoff
Transition refusal	Approved, exploratory, or refused targeted-transition disposition	Restrict or reject the transition with reasons
Technical failure	FAILED_TECHNICAL and technical failure class	Diagnose execution and apply the declared rerun policy
Reproducibility failure	FAILED_REPRODUCIBILITY	Review dispersion, batches, protocol, and repeats
Biological failure	FAILED_BIOLOGICAL	Retain as interpretable adverse evidence; do not relabel as technical
Inconclusive outcome	INCONCLUSIVE with uncertainty and caveats	Hold promotion or request resolving work
Promotion conflict	Blocked promotion state or audit issue	Preserve the observation and resolve the promotion gate

flowchart TD
    P[Planned work] --> V{Plan and readiness valid?}
    V -->|no| B[Blocked with named findings]
    V -->|yes| O[Observe]
    O --> Q{QC and interpretation}
    Q -->|technical| T[Rerun or repair]
    Q -->|biological| E[Adverse evidence]
    Q -->|inconclusive| H[Hold and resolve]
    Q -->|supported| R[Promotion review]

A failed acceptance rule does not identify the failure class by itself. QC, replicate behavior, controls, material state, censoring, and protocol context determine whether the correct response is rerun, redesign, adverse-evidence promotion, or refusal to interpret.

Build a failure-classification packet

Evidence	Question it answers
approved plan and handoff	what question, protocol, controls, materials, and acceptance criteria were authorized?
custody and execution record	which samples, batches, instruments, operators, timings, and deviations belong to the observation?
raw and processed measurements	what was observed before and after declared transformation?
control and QC outcomes	did positive, negative, process, reference, and system controls behave as required?
replicate and batch behavior	is the result repeatable, dispersed, censored, or associated with one batch?
material and protocol state	were identity, integrity, storage, preparation, and procedural requirements satisfied?
detection and acceptance limits	was absence distinguishable from below-detection, saturation, or an invalid measurement?
classification and authority	who assigned technical, reproducibility, biological, inconclusive, or uninterpretable status, and under which policy?

The packet supports one response: rerun unchanged, repair and rerun, redesign, accept adverse biological evidence, hold, or refuse interpretation. A rerun is not the default for biological failure, and promotion is not available while technical validity or provenance remains unresolved.