Net-vs-Gross Reconciliation

Every weighment on the floor produces three legitimate quantities that must be reconciled: GROSS (what the balance saw on the pan, including container + scoop + lid + any tooling left on the pan), TARE (the certified non-product mass — container with lid, or empty boat — established under the tare-verified weighing chain), and NET (gross − tare, the formula-relevant mass that goes into the batch). The fourth quantity, ISSUED (mass picked from the warehouse against the WO pick list), must close back to the sum of nets + the documented residual returned to inventory or written off.

The §211.188(b)(11) batch record requirement is unambiguous: 'a statement of the actual yield and a statement of the percentage of theoretical yield' AND 'a complete record of all stages of processing'. The 'all stages' phrase carries the audit hook: a BMR cell that records only the net (e.g. '125.0 g charged') without the gross + tare context cannot be reconstructed, cannot be challenged, and cannot be verified. The first inspector question on any low-yield investigation is 'show me the gross reading and the tare for charge 7 on this batch' — if the answer is 'we only logged the net', the §211.188 finding writes itself.

Every gram of the issued-vs-charged delta must map to one of six authorised loss categories. 'Unaccounted' is not a category; it is a §211.192 finding. The categories must be pre-validated during PPQ + characterised with an expected range during Stage-3 CPV.

1. Operator scans the issued container (with lot + tare cert + remaining mass) at the balance; engine records the issued lot + the pre-charge container mass.
2. Tare event: empty weigh-boat placed on pan; balance auto-tares; engine writes the tare event with mass + UTC + balance + sigma fingerprint per the tare-verified weighing chain.
3. Charge event: operator transfers powder; engine captures gross when stability + dwell pass; computes net = gross − tare; compares to post-adjustment target + tolerance band.
4. Container re-weigh: issued container placed on a second balance (or same balance with full re-tare); engine computes mass-removed = pre-charge − post-charge; reconciles mass-removed = net + container_residue + spill_estimate.
5. Engine writes a single dispense_result row with eight fields (issued_lot, pre_charge_container_mass, tare, gross, net, post_charge_container_mass, residue_estimate, spill_estimate) + per-field e-sig + auto-flags any mass-removed vs net delta outside the authorised-loss envelope.
6. Tooling reconciliation: shared scoops / spatulas auto-flagged for carry-over re-weigh at line-clearance + shift-end; carry-over written to a per-tool ledger that genealogy-links to the charges the tool touched.
7. Per-component running ledger: engine maintains live issued-vs-charged delta per component per batch; in-batch breach of the envelope pages the QA-on-call BEFORE the next charge can start.
8. Sample + IPC retention: any mass drawn for testing writes to a separate sample ledger; sample mass is not counted toward charge but IS counted toward issued reconciliation.
9. Returned-to-inventory: residual mass returned to source lot requires a paired inventory transaction with destination + container + chain-of-custody + supervisor e-sig.
10. End-of-charge close-out: operator e-signs the per-charge reconciliation; engine commits dispense_result + per-component delta + loss-category breakdown; BMR cell renders with full provenance.

Per-component reconciliation must be LIVE across the batch, not aggregated retrospectively. A 12-charge multi-component batch maintains a running ledger of issued + charged + residue + sample + returned + scrap per component, updated after each weighment, with hard-block triggers when any component's running delta breaches the validated envelope. The ledger is rendered on the BMR + queryable from the kiosk + visible to the supervisor at any time during the batch.

The 'Status' column carries the decision logic: in-envelope (green) + investigate (amber) + breach (red, batch hard-stops). The amber 'scrap-flagged: investigate' state allows the batch to continue while QA reviews the scrap event in parallel; a red 'breach' state hard-stops the next charge until QA dispositions the gap. The hard-stop is RLS-enforced at the database — no service-role-key bypass.

1. Only NET is logged — BMR cell shows '125.0 g of API-A charged'; no gross + tare provenance; reconstruction impossible; §211.188 + ALCOA+ accuracy deficiency on the first inspector question.
2. End-of-batch aggregation only — per-charge data not retained; only batch-level 'issued vs charged' summary; over-charges + under-charges silently cancel; PV Stage-3 CPV trending impossible.
3. 'Unaccounted' loss category in use — BMR shows 1.2% loss as 'unaccounted'; not mapped to a validated category; §211.192 finding because every gram must live in an authorised category.
4. Authorised-loss envelope copied from a similar product — no PPQ derivation, no ICH Q9 QRM justification; envelope is folklore; first OOS yield investigation finds the envelope cannot be defended.
5. Container residue estimated, not weighed — operator visually estimates 'about 0.5%' of residue per charge; no actual container re-weigh; estimate is systematic + invisible bias; cannot support root-cause investigation.
6. Returned-to-inventory without paired transaction — residual mass returned to source container; no inventory transaction; warehouse-side genealogy breaks; next WO using same lot has incorrect issued-mass baseline.
7. Shared tooling carry-over ignored — same scoop used for 5 different components across a shift; carry-over of 0.1% per charge accumulates to cross-contamination + 1% mass-balance error; not tracked because tool ledger doesn't exist.
8. Mid-batch breach 'reviewed and accepted' without deviation — running ledger shows component-B at 1.5% over loss envelope after charge 8; operator + supervisor agree to 'continue and investigate at close'; no deviation opened; closure becomes retroactive; §211.192 + 211.100 finding identical to the 'continue and document' SOP gap.

• Per-charge gross+tare+net capture rate — fraction of weighments with all three values logged + reconciled (target 100%); any gap is a §211.188 finding.
• Authorised-loss category coverage — fraction of issued-vs-charged delta mapped to an authorised category (target 100%); any 'unaccounted' is a §211.192 finding.
• In-envelope per-charge rate — fraction of charges where the mass-removed-vs-net delta is inside the validated envelope (target ≥95%); chronic breaches trigger SPC investigation.
• Live ledger breach response time — median time from running-ledger amber/red state to QA disposition (target ≤30 min for amber, ≤10 min for red).
• Tooling carry-over within envelope — fraction of shared-tool shifts with documented carry-over inside envelope (target ≥98%); chronic breaches trigger dedicated-tool revisit.
• Returned-to-inventory paired-transaction rate — fraction of residual returns with paired inventory transaction (target 100%); any unpaired return is genealogy breakage.
• Stage-3 CPV reconciliation trending coverage — fraction of products with per-charge per-component reconciliation trends visible on the quarterly review (target 100%); gaps are 2024 FDA PV finding-bait.
• Container-residue measured-vs-estimated ratio — fraction of charges where residue is measured by re-weigh (target ≥95%); chronic estimation is a data-integrity gap.

1. MMR carries authorised_loss_categories[] per component with per-category envelope_min_pct + envelope_max_pct + ppq_source + qrm_doc_ref + change_control_id; envelopes are change-controlled and cannot be edited without two-person e-sig + QA approval.
2. Dispense_result schema enforces NOT NULL on issued_lot + pre_charge_container_mass + tare + gross + net + post_charge_container_mass + residue_estimate + spill_estimate; RLS rejects writes missing any field — service-role-key not exempt.
3. Engine computes mass-removed = pre_charge − post_charge; auto-reconciles mass-removed = net + residue + spill; flags any |delta| > 0.01 g or > validated tolerance whichever is tighter as an investigation candidate.
4. Per-component running ledger updates after every charge; ledger state (green / amber / red) is denormalised on the batch row for fast kiosk + supervisor dashboard queries.
5. Amber state surfaces a kiosk banner + emails QA-on-call; batch continues. Red state hard-blocks the next charge via RLS until a QA deviation_id is attached to the ledger row.
6. Shared-tool ledger tracks every tool's per-charge association + carry-over re-weigh at line-clearance + shift-end; tool-level carry-over outside envelope auto-opens a deviation + flags affected charges in genealogy.
7. Returned-to-inventory workflow requires a destination lot + container + supervisor e-sig + paired inventory_transaction row; orphan returns hard-blocked.
8. BMR rendering: every charge cell renders gross + tare + net + container_before + container_after + residue + spill + sample + scrap + per-component running delta — full provenance, no aggregation hiding.
9. §211.192 review template auto-attaches per-charge reconciliation breakdown + per-component running ledger + every amber/red event + every deviation; reviewer must explicitly disposition any out-of-envelope event before closing.
10. Quarterly ICH Q10 §3.2.5 + FDA PV Stage-3 CPV review aggregates per-charge per-component reconciliation trends with control charts per-product per-component; sustained drift triggers envelope re-validation + PPQ rationale revisit + change-control update.

• Q: Can we use a single 'process loss' category to cover container residue + spill + dust? A: No — each contributes a different root cause + remediation + risk profile. Container residue is a container-design + cleaning + powder-cohesion issue; spill is an operator-technique + isolator-design issue; dust is a ventilation + powder-particle-size issue. ICH Q9(R1) requires categories granular enough to drive distinct CAPA decisions.
• Q: How often must authorised-loss envelopes be re-validated? A: At least at quarterly product review (ICH Q10 §3.2.5) for trend assessment, and formally re-validated whenever PPQ-bracket conditions change: new container supplier, new powder grade, new operator group + training change, new isolator + ventilation system, new scoop / tooling material. Sustained Stage-3 CPV drift triggers ad-hoc re-validation.
• Q: Do we need container re-weigh for every charge or can we statistically sample? A: Statistical sampling is acceptable for low-risk + wide-envelope components where PPQ data supports the sampling plan + QRM documents the rationale. High-potency APIs + low-mass charges + tight tolerances require 100% container re-weigh. The decision is per-component + change-controlled, not an SOP-wide default.
• Q: What about liquids — gross / tare / net by volume vs mass? A: Always mass for reconciliation. Volume measurements have a density + temperature dependency that introduces uncontrolled error; mass-based reconciliation is the regulatory expectation. Convert to volume only for downstream process-specification purposes, not for §211.188(b)(11) reconciliation.
• Q: How do we handle reconciliation for continuous processes (continuous granulation, continuous tableting)? A: Continuous processes use loss-in-weight feeders + check-weighers + start/end mass-balance reconciliation per ICH Q13. The per-charge model is replaced by per-time-window reconciliation (e.g. per hour) with the same six-category loss breakdown + live-ledger discipline.
• Q: Can returned-to-inventory close a per-component running ledger amber state? A: Yes if (a) the return is performed with paired inventory transaction + supervisor e-sig, (b) the returned mass is traceable to a specific over-issue earlier in the batch, and (c) the post-return running delta is back inside envelope. The system writes the return event into the ledger; the amber state clears automatically; no separate deviation is required.
• Q: Do we need to reconcile retain samples + IPC samples against issued? A: Yes — the issued mass must close to: charged + container_residue + spill + sample + retained + returned + scrap. Sample mass is part of the reconciliation even though it is not part of the batch charge. Failing to count sample mass produces a 'phantom loss' equal to sample mass, often misdiagnosed as container residue.