Outcome Learning Loops

Outcome loops compare requested follow-up with observed work and carry the difference into the next recommendation. The prior recommendation, approved handoff, and observation remain separate immutable records.

What One Loop Tells You

Each workflow-family loop keeps the same questions visible:

• what the recommendation posture was before follow-up
• what assays were requested
• what assays actually happened
• whether the loop was worth the assay spend
• how the observed result should narrow or strengthen the next recommendation

flowchart LR
    request["requested assays and expected information"] --> observation["observed assays, QC, and deviations"]
    observation --> reconcile["match, block, weaken, or contradict"]
    reconcile --> worth{"information justified burden?"}
    worth -->|yes| revise["versioned evidence and decision revision"]
    worth -->|no| narrow["narrow, hold, or refuse"]
    revise --> preserve["retain prior records for calibration"]
    narrow --> preserve

Cross-Family Snapshot

workflow family	initial posture	revised posture	worth the assay spend	current lesson
dda	recommend_with_downgrade	recommend_with_downgrade	yes	matched assays preserved the intended closure loop without widening public language
dia	recommend_with_downgrade	do_not_recommend	no	requested assay loss should block any attempt to treat DIA follow-up as laboratory closure
lfq	do_not_recommend	do_not_recommend	no	low-information repeats should stay refused when the biological conclusion is still design-limited
multiplex	do_not_recommend	do_not_recommend	no	no shipped requested-versus-observed outcome loop exists for this family yet
ptm	do_not_recommend	recommend_with_downgrade	yes	site-level ambiguity can still repay follow-up when the closure question is narrow and explicitly targetable
targeted	do_not_recommend	recommend_with_downgrade	yes	fast calibration-facing loops can be worth it even when broader targeted authority remains bounded

dda

• initial posture: recommend_with_downgrade
• revised posture after outcome: recommend_with_downgrade
• worth the assay spend: yes
• requested assays: dda-repeat-digest, dda-pooled-reference, dda-carryover-blank
• observed assays: dda-repeat-digest, dda-pooled-reference, dda-carryover-blank
• matched assays: dda-repeat-digest, dda-pooled-reference, dda-carryover-blank
• blocked assays: none
• weakened assays: none

learning points

• matched assays preserved the intended closure loop without widening public language
• the follow-up repaid cost because it stabilized the contaminant and pooled-reference boundary

next adjustments

• matched assays preserved the intended closure loop without widening public language
• the follow-up repaid cost because it stabilized the contaminant and pooled-reference boundary

evidence paths

• artifacts/lab/flagship-follow-up-outcomes/dda.json
• claim:dda_target_decoy_stability

dia

• initial posture: recommend_with_downgrade
• revised posture after outcome: do_not_recommend
• worth the assay spend: no
• requested assays: dia-library-bridge, dia-matrix-shift-repeat
• observed assays: dia-matrix-shift-repeat
• matched assays: dia-matrix-shift-repeat
• blocked assays: dia-library-bridge
• weakened assays: dia-matrix-shift-repeat

learning points

• requested assay loss should block any attempt to treat DIA follow-up as laboratory closure
• strong import-backed review does not rescue a loop that fails its own library bridge

next adjustments

• requested assay loss should block any attempt to treat DIA follow-up as laboratory closure
• strong import-backed review does not rescue a loop that fails its own library bridge
• feed the revised follow-up result back into future recommendation posture instead of keeping the original recommendation sentence unchanged
• treat blocked assays as explicit evidence for narrower or slower future follow-up
• carry weakened assays forward as downgrade evidence rather than as partial confirmation

evidence paths

• artifacts/lab/flagship-follow-up-outcomes/dia.json

lfq

• initial posture: do_not_recommend
• revised posture after outcome: do_not_recommend
• worth the assay spend: no
• requested assays: lfq-extra-replicate-block, lfq-randomized-repeat
• observed assays: lfq-extra-replicate-block
• matched assays: lfq-extra-replicate-block
• blocked assays: lfq-randomized-repeat
• weakened assays: lfq-extra-replicate-block

learning points

• low-information repeats should stay refused when the biological conclusion is still design-limited

next adjustments

• low-information repeats should stay refused when the biological conclusion is still design-limited
• treat blocked assays as explicit evidence for narrower or slower future follow-up
• carry weakened assays forward as downgrade evidence rather than as partial confirmation

evidence paths

• artifacts/lab/flagship-follow-up-outcomes/lfq.json

multiplex

• initial posture: do_not_recommend
• revised posture after outcome: do_not_recommend
• worth the assay spend: no
• requested assays: none recorded
• observed assays: none recorded
• matched assays: none
• blocked assays: none
• weakened assays: none

learning points

• no shipped requested-versus-observed outcome loop exists for this family yet

next adjustments

• publish a dedicated downstream lab consequence and observed outcome loop before strengthening recommendation posture

evidence paths

• artifacts/intelligence/recommendation-packets/multiplex.json

ptm

• initial posture: do_not_recommend
• revised posture after outcome: recommend_with_downgrade
• worth the assay spend: yes
• requested assays: ptm-site-targetability, ptm-orthogonal-validation
• observed assays: ptm-site-targetability, ptm-orthogonal-validation
• matched assays: ptm-site-targetability, ptm-orthogonal-validation
• blocked assays: none
• weakened assays: ptm-site-targetability

learning points

• site-level ambiguity can still repay follow-up when the closure question is narrow and explicitly targetable
• useful PTM follow-up should strengthen a bounded claim rather than erase the wider ambiguity warning

next adjustments

• site-level ambiguity can still repay follow-up when the closure question is narrow and explicitly targetable
• useful PTM follow-up should strengthen a bounded claim rather than erase the wider ambiguity warning
• feed the revised follow-up result back into future recommendation posture instead of keeping the original recommendation sentence unchanged
• carry weakened assays forward as downgrade evidence rather than as partial confirmation

evidence paths

• artifacts/lab/flagship-follow-up-outcomes/ptm.json
• claim:ptm_site_resolution_boundary

targeted

• initial posture: do_not_recommend
• revised posture after outcome: recommend_with_downgrade
• worth the assay spend: yes
• requested assays: prm-assay, orthogonal-assay
• observed assays: prm-assay, orthogonal-assay
• matched assays: prm-assay, orthogonal-assay
• blocked assays: none
• weakened assays: none

learning points

• fast calibration-facing loops can be worth it even when broader targeted authority remains bounded
• follow-up value here comes from interference clarification, not from pretending vendor-parity authority exists

next adjustments

• fast calibration-facing loops can be worth it even when broader targeted authority remains bounded
• follow-up value here comes from interference clarification, not from pretending vendor-parity authority exists
• feed the revised follow-up result back into future recommendation posture instead of keeping the original recommendation sentence unchanged

evidence paths

• artifacts/lab/flagship-follow-up-outcomes/targeted.json
• claim:targeted_transition_consistency
• claim:targeted_interference_boundary

Boundary

An outcome loop can revise the next recommendation only at the scope supported by its controls and observed assays. Missing assays, failed controls, and weak information gain are evidence for narrowing; they are not silently discarded operational details.