Theoretical vs Actual Yield %

Theoretical vs Actual Yield % is the end-of-batch closure ratio: actual finished output ÷ scaled theoretical × 100. It is the headline number on the BMR closure page, the trigger for the 21 CFR 211.192 production record review, the input the §211.110(a)(4) percentage-of-yield acceptance limit is tested against, and (for supplements) the trigger for the §111.310 yield-deviation investigation. Every regulated batch record must include both the actual yield and the percentage of theoretical.

It is distinct from in-process reconciliation, which closes mass / count / volume balances at each stage gate so root cause stays local. Yield % is the whole-batch closure number — typically rolled up only after packaging — and its job is to test whether the batch landed inside its validated yield envelope, not to localise the loss. The two are complementary: stage-gate reconciliation tells you where loss happened; yield % tells you whether the cumulative result is still in spec.

yield% = (actual_net ÷ theoretical_scaled) × 100, where actual_net = actual_gross − IPC samples − retain samples − authorised loss (validated equipment hold-up, validated dust extraction, line-clearance discards, weight-rejected units). The scaled theoretical is the yield-adjusted-batch-size figure locked into the WO snapshot at release, not the nominal MMR.

The validated yield envelope is the PPQ-derived range of yield % the process is qualified to operate within. Typical envelopes: 95–102% for solid-dose; 97–101% for sterile liquids; 90–105% for early API steps where overage and impurity removal widen the band. The envelope is symmetric around 100% by intent — the §211.110(a)(4) wording is 'percentages of theoretical yield … beyond which deviation from the percentage of theoretical yield must be investigated.' Beyond, not below.

• Low-yield breach (yield < lower limit) — opens a deviation under §211.192 + §111.310. Investigation must identify the lost mass (in-process reconciliation stage-Δ data is the first artefact) and either explain it (validated authorised-loss undercount, equipment fault, container failure) or escalate to CAPA + impact assessment for already-released batches sharing the root cause.
• High-yield breach (yield > upper limit) — also opens a deviation. The bias is rarely 'we made extra' — far more often it is a measurement error (scale calibration drift, fill weight off-target high, missing IPC-sample subtraction) or, more seriously, an incorrect MMR theoretical (water-of-crystallisation not accounted for, density figure stale, basis-conversion missing — pointing back to a SBF / LOD / PF error upstream). High-yield investigations are some of the most data-integrity-sensitive in GMP.
• Right-at-the-edge (e.g. 101.9% against 95–102% envelope) — not a breach but a CPV warning. The Stage-3 CPV chart should already show this as a Western-Electric rule 1 candidate; trending action is required even though the single-point disposition is pass.
• Many-batches-clustered-just-inside-the-edge — the envelope is wrong (set too generous) or the process is drifting. Quarterly ICH Q10 §3.2.5 PQS review surfaces this; CPV charts make it visible per batch.

Every batch — pass or fail — closes with a §211.192 production record review by independent QA. The review is not a rubber stamp: it confirms the calculations are correct, the actual-net was derived properly, the denominator was the scaled theoretical, the authorised-loss schedule reconciles to the stage-gate evidence, and the disposition decision is supported. For an out-of-envelope batch the review is more rigorous and may not close until the deviation is closed under §211.192 + §211.100.

• Reviewer is an independent QA function distinct from the production unit that ran the batch — required by §211.192.
• Review e-sig is Part 11 (link to user, link to record, link to timestamp, link to meaning — 'production record review approved — disposition: release / quarantine / reject').
• If the review identifies a calculation error, the disposition cannot be approved; the BMR returns to production for correction, which itself requires a recalculation e-sig (see /glossary/recalculation-e-signature).
• Disposition decision (release / quarantine / reject) is captured with the review e-sig; release is then a separate e-sig event by QA leadership for high-stakes batches per the local SOP.

1. Denominator = nominal MMR on a scaled batch — every yield-adjusted batch reads as massively under-yielded; the deviation log fills up; eventually production complains and the envelope is silently widened, hiding genuine low-yield drift behind the noise.
2. Numerator = gross dispensed mass (IPC samples ignored) — apparent yield is systematically high by the sample mass; subtle low-yield drift in the actual process is masked because the IPC subtraction is doing the compensating.
3. Out-of-envelope batch closed with 'process loss within typical range' as the deviation conclusion — no root cause, no CAPA, the same loss recurs; the most-cited GMP deviation-closure 483 of the past decade.
4. Envelope widened after repeated failures rather than process investigated — change control raised to expand the envelope from 95–102% to 90–104% to 'reflect real-world variability'; FDA position is that the envelope must be PPQ-justified, not failure-justified.
5. Cpk drift unseen by point-yield checks — every batch passes against the envelope but the Cpk has fallen from 1.65 to 0.95 over 18 months; point-yield checks miss this; only Stage-3 CPV trending under ICH Q10 §3.2.5 surfaces the loss of capability.
6. High-yield investigation skipped — 103% yield closed without investigation on the assumption that 'yield over 100% is good news'; the actual root cause is a stale density figure on the MMR; subsequent batches with corrected density show systematic under-fill against label claim.

• Yield % within validated envelope (per product, per quarter) — fraction of batches inside their validated envelope (target ≥95%; below triggers process investigation, not envelope widening).
• Rolling 12-batch Cpk for yield % (per product) — process-capability index tracked under ICH Q10 §3.2.5; target ≥1.33; trending below 1.33 triggers CPV action even if individual batches pass.
• Yield-deviation closure quality % — fraction of yield-deviation investigations closed with an identified root cause + impact assessment + CAPA where indicated (target 100%; 'process loss within typical range' counts as zero).
• Time-to-close yield deviation (median days) — target ≤14 days; longer than 30 days is a backlog signal.
• High-yield investigation rate — fraction of >upper-limit batches that ran through full investigation (target 100%); a low rate is a data-integrity yellow flag (high yields being waved through).
• Masked-low-yield detection (audit-derived) — number of batches where numerator-vs-denominator review during periodic audit identified a hidden yield problem (any non-zero number is a root-cause action).
• Envelope-widening events (per product per year) — change-control events that expanded a yield envelope; target ≤1; >1 in a year is a process-investigation trigger, not a paperwork action.

1. Denominator from snapshot: yield % is always computed against work_orders.mmr_snapshot.scaled_theoretical, never against the live MMR or the nominal figure; the snapshot was locked at WO release per the YA contract.
2. Numerator from in-process reconciliation: actual_net is derived by subtracting the authorised-loss schedule (IPC samples + retain samples + validated hold-up + validated dust extraction + line-clearance discards + weight-rejected units) from actual_gross; every subtraction must have its evidence record present or the BMR will not close.
3. Live yield-% render on BMR closure page: nominal MMR figure, scaled theoretical, scale factor, actual gross, itemised authorised loss, actual net, yield %, validated envelope, and pass/fail badge — all shown side-by-side so the §211.192 reviewer can verify the math in one screen.
4. Out-of-envelope hard-block: BMR cannot move to disposition until either yield % is inside the envelope or a deviation has been opened against the batch under §211.192 + §211.100. Closure of the deviation is required before disposition can be release.
5. Reviewer-distinct-from-preparer enforcement: §211.192 reviewer dropdown excludes any user with a preparer e-sig event on the same batch; database constraint prevents bypass.
6. Part 11 review e-sig: link to user / record / timestamp / meaning ('production record review approved — disposition: release / quarantine / reject'); stored alongside the snapshot in report_renders for forensic traceability.
7. Stage-3 CPV trending: per-product yield % is fed into the CPV chart on the Quality dashboard with Western-Electric + Nelson rules; out-of-control points page the process-engineering team independent of single-batch disposition.
8. Quarterly ICH Q10 §3.2.5 PQS management review: per-product yield distribution, Cpk trend, deviation rate, deviation closure quality, envelope-widening events, and masked-low-yield audit findings — surfaced as a single report-pack the QA director presents at the management review.
9. Recalculation guardrail: any post-close recalculation of the yield numerator or denominator (e.g. an authorised-loss figure corrected on appeal) requires the recalculation e-sig contract (/glossary/recalculation-e-signature) — original values preserved in the audit trail, recalculated values rendered alongside, both linked into report_renders.
10. Customer-facing CoA: yield % is not on the CoA (it is a manufacturing-internal acceptance criterion, not a release spec for the customer) but the §211.192 review approval state gates CoA generation — no review approval, no CoA.