Yield Reconciliation

The accounting equation:

• Theoretical yield (per MMR, scaled to batch size) = Actual yield + Sampling loss + Line loss + Scrap + Rework + Quarantine + Unaccounted
• If "Unaccounted" is non-zero outside the permissible variance band, 211.192 mandates an investigation.

Every term in the equation must be measurable, traceable and defensible:

• Sampling loss — the mass of in-process samples taken for IPC + QC release tests; recorded at sample event.
• Line loss — process equipment hold-up, transfer losses, residue in vessels / lines / filter housings; estimated from validated process characterisation.
• Scrap — physically discarded material; weighed and entered into the scrap account.
• Rework — material reprocessed back into the batch or another batch under formal rework approval.
• Quarantine — material held pending disposition; reconciles at disposition time.
• Unaccounted — what cannot be explained. Greater than zero outside the variance band → investigation.

211.186(b)(7) requires the MMR to state "theoretical yield, including the maximum and minimum percentages of theoretical yield beyond which investigation is required". Typical bands depend on product complexity:

• Compressed tablets / capsules — typically ± 2-3 % of theoretical for an established process.
• Liquid bulk — ± 1-2 %.
• Sterile filtration / fill — ± 3-5 % (filter hold-up + line clearance).
• Lyophilised products — ± 3-5 %.
• Solid-dose API + intermediates — ± 5 % for early-stage; tighter as the process matures.

The band is established from validation data + process capability and tightened as Cpk improves. It is part of the recipe, frozen into the WO snapshot at release.

Reconciliation at the end of the batch tells you something went wrong; reconciliation at every step tells you where. EU GMP Chapter 5 + FDA guidance both expect reconciliation at every critical step (dispense, granulation, blending, compression, coating, fill). In V5 this is the in-process reconciliation pattern:

• After dispense — reconcile dispensed mass against MMR target ± tolerance for every component.
• After granulation — reconcile granulation output against dispensed input ± documented losses (sieve overs, container hold-up).
• After blending — reconcile blender discharge against blender charge ± hold-up.
• After compression / encapsulation — reconcile tablet / capsule count against expected (mass / unit weight).
• After coating — reconcile coated mass against uncoated mass + coating suspension applied.
• After packaging — reconcile packaged units against bulk released.

An out-of-band step-level reconciliation surfaces immediately, before subsequent steps add complexity.

Mass measurements at every transfer must reconcile gross (container + content) minus tare (container) to net (content). Common pitfalls:

• Tare value not recorded at the same scale event as gross — drift produces a phantom variance.
• Container weight assumed ("standard pail = 1.2 kg") rather than measured per use.
• Tare lost between dispense and final reconciliation; gross-only reading cannot reconcile.
• Multiple-container dispenses summed gross then tared once — arithmetic error.

V5 enforces tare-verified weighing: every weighing records gross, tare and net at the same timestamp on the same scale, with the scale's USP <41> minimum weight check applied.

211.192 says: "Any unexplained discrepancy ... or the failure of a batch or any of its components to meet any of its specifications shall be thoroughly investigated ... The investigation shall extend to other batches of the same drug product and other drug products that may have been associated with the specific failure or discrepancy."

• Unexplained yield gap > permissible variance — formal deviation, RCA, CAPA evaluation, batch disposition decision.
• Repeat trend (same step, same product, gap creeping) — even within band — flagged for management review under ICH Q10.
• Cross-batch impact assessment — if a process change caused the gap, sister batches may be affected.

A within-band yield variance with documented loss accounting is a normal release. An out-of-band variance is not automatically a reject — the QA may release if the investigation concludes the variance does not affect identity / strength / quality / purity. The investigation conclusion and the QA release decision land in the BMR.

• Permissible variance band set wide enough to accommodate every batch — defeats the investigation trigger.
• Reconciliation done only at final step — root cause cannot be localised.
• Line loss treated as an unmeasured catch-all — auditors will challenge.
• Sampling loss not recorded at sample event; calculated after the fact.
• Scrap weighed in pounds when MMR is in kilograms — unit-error variance.
• Rework absorbed without formal rework approval — Part 11 + 211.192 violation.
• Tare assumed not measured — phantom variance.
• Variance acknowledged but no deviation raised — 211.192 violation, almost always cited in FDA 483s.