As-Found / As-Left Calibration

As-found / as-left calibration is two independent measurement sets captured at every calibration event against the same traceable reference. As-found is the instrument's reading BEFORE any adjustment, cleaning, levelling, or service — it is the drift evidence, the answer to 'what was this instrument telling the floor between the last calibration and today?'. As-left is the reading AFTER any adjustment, against the same reference, in the same configuration the instrument will be returned to service in. It is the forward-confidence evidence, the answer to 'what will this instrument tell the floor from today forward?'. The two sets are not interchangeable, not optional, and not a vendor-template convenience — they are separately consumed by separate regulatory questions and separately audited at every inspection.

The discipline appears in §211.68(a) ('Automatic, mechanical, or electronic equipment ... that will perform a function satisfactorily, may be used in the manufacture, processing, packing, and holding of a drug product. If such equipment is so used, it shall be routinely calibrated, inspected, or checked according to a written program designed to assure proper performance'), §211.160(b)(4) (calibration accuracy appropriate to the specifications being judged), 820.72(a) ('Each manufacturer shall ensure that all inspection, measuring, and test equipment ... is suitable for its intended purposes and is capable of producing valid results'), EU GMP Chapter 3 §3.41, Annex 15 §3, ISO/IEC 17025 §6.4.6, and ISO 10012 §7.3. None of those clauses uses the literal phrase 'as-found / as-left' — but every inspector reads them as requiring it, because without as-found there is no answer to the back-impact question.

A defensible calibration program runs TWO tolerance widths against the same test-points: a wider as-found 'in-use' tolerance and a tighter as-left 'post-service' tolerance. The as-found tolerance is the threshold below which routine drift is acceptable (no batch impact, no deviation); the as-left tolerance is the post-service confidence band that the next as-found cycle is judged against. The gap between the two is the operational headroom — it is what allows the instrument to drift gradually between calibrations without immediately triggering OOT.

An as-found out-of-tolerance (OOT) result is never just 'recalibrate and continue'. It is the trigger for a §211.192 + §211.100 + 820.100 back-impact assessment of every batch produced since the previous as-left passed. The chain is deterministic, sequential, and audit-trailed end-to-end.

1. Capture the as-found reading + flag OOT in the calibration record + lock the instrument (kiosk hard-block, no dispense_result writes accepted) — instrument status moves to 'out-of-service pending investigation'.
2. Auto-open a deviation with the OOT delta, the test-point, the reference cert, the operator, and a back-impact-required flag pre-populated.
3. Auto-query every batch / WO / dispense_result that used this instrument since the previous as-left (the calibration_valid_until window) — produce a list with batch IDs, products, dates, operators, dispense counts, and dependent QC results.
4. QRM-score each affected batch per ICH Q9(R1): is the OOT delta material to the charge tolerance the instrument was enforcing? Inside the operating-tolerance band → low risk; just outside → medium; multiple-fold outside → high.
5. For low-risk batches: document the assessment, sign off, no batch action required. For medium-risk: trigger §211.110 IPC re-test + finished-product re-assay where samples remain. For high-risk: quarantine + investigate + potentially recall.
6. Capture the back-impact assessment as a §211.192 production-record-review artefact for every affected batch — the calibration deviation is now part of each batch's BMR closure trail.
7. Service the instrument; perform the adjustment; capture as-left; pass as-left tolerance; return to service; instrument status moves back to 'in-service' only after the deviation back-impact assessment is closed.
8. Quarterly aggregate of OOT events + back-impact extents + batch dispositions feeds the ICH Q10 §3.2.5 product review — chronic OOT on a single instrument is a capability-ratio + PM-frequency + environment trigger, not a calibration-frequency issue alone.

1. Calibration due-date arrives; instrument status flips to 'calibration-due' but stays in-service until performed (with a soft-warning + a 7-day hard-block grace window).
2. Calibration technician opens the calibration record in V5; system pre-populates instrument ID + last as-left + tolerance bands + reference cert IDs available + test-point plan from the calibration SOP.
3. Technician confirms reference identity + cert valid-through + uncertainty; mismatch (expired cert, mismatched uncertainty, untraceable reference) is a hard-block.
4. Technician captures as-found at each test-point (low / mid / high + replicate as defined); system evaluates against as-found tolerance; OOT auto-opens deviation + back-impact workflow.
5. Technician performs adjustment (zero, span, level, clean, service) per SOP; adjustment captured as a separate audited event with what + by-whom + authority.
6. Technician captures as-left at each test-point against the same reference; system evaluates against as-left tolerance; fail = re-adjust + re-test, repeat until pass or service-escalation.
7. Calibration technician e-signs the as-found + as-left + adjustment chain; reviewer (independent role) e-signs the calibration record; instrument status flips to 'in-service'; calibration_valid_until + next-cal-due dates written to instrument master.
8. Audit trail captures every measurement, adjustment, signature, and disposition with UTC timestamp; original readings are immutable; corrections (rare, audit-trailed) are e-sig'd separately.
9. Calibration PDF generated + stored in the regulated-reports bucket; cross-referenced from the instrument master + from every dispense_result that consumed the prior as-left.
10. If as-found was OOT: back-impact deviation drives §211.192 batch-by-batch review; affected batches' BMRs are amended with the calibration deviation reference; QA disposition is recorded against each.

1. Pass/fail-only template — vendor calibration certificate captures 'pass' for each test-point with no raw as-found / as-left readings; first inspector question 'what was the reading before you adjusted it?' has no answer; §211.68 + §211.194 + Annex 11 §9 finding.
2. As-left written, as-found never captured — calibration log shows only post-adjustment readings; drift evidence missing; back-impact assessment impossible; data-integrity finding under MHRA DI 2018.
3. OOT discovered, instrument adjusted, no deviation opened, no back-impact — technician 'corrected' the OOT and returned to service; affected batches produced over the prior calibration interval never assessed; §211.192 + §211.100 finding the first time an inspector pulls the calibration log alongside the batch list.
4. Back-impact 'all batches OK' with no QRM evidence — assessor signs off batch impact as null without per-batch QRM scoring; ICH Q9 + §211.192 deficiency; back-impact must be evidence-based per batch.
5. Calibration vendor adjusts off-site, returns instrument with as-left only — vendor cannot or does not capture as-found before adjusting at their facility; site has no drift evidence; vendor service contract is non-compliant; common with shared analytical instruments returned to OEM for service.
6. Reference standard uncertainty omitted — calibration record cites reference ID but not cert + uncertainty; traceability chain incomplete; ISO/IEC 17025 §6.5 deficiency.
7. Adjustment-required threshold ignored — instrument drifts to 70% of in-use tolerance every cycle; technician 'in-tolerance, no adjustment'; next cycle OOT; preventable OOT becomes back-impact event; SPC discipline missing.
8. Tolerance widened reactively to clear an OOT — instrument repeatedly OOT against original tolerance; tolerance silently widened to make it pass; original tolerance was tied to the operating spec; widening tolerance breaks the capability-ratio + spec-tie-back chain; §211.100 + ICH Q9 + change-control deficiency.

• As-found OOT rate — fraction of calibration events with as-found out-of-tolerance; target ≤2%; sustained higher rate indicates calibration-frequency-too-low or environment-degraded or instrument-end-of-life.
• As-left first-pass rate — fraction of calibration events where as-left passed on first attempt without re-adjustment; target ≥95%; lower indicates technique gap or instrument health concern.
• Back-impact assessment cycle time — median hours from as-found OOT capture to deviation closure with back-impact disposition; target ≤72h; longer delays quarantine downstream batches.
• Calibration-due adherence rate — fraction of calibration events performed on or before due date; target ≥98%; misses force kiosk hard-block + WO deferrals.
• Adjustment-required-threshold capture rate — fraction of in-tolerance-but-approaching events that triggered preventive adjustment; target captures the SPC discipline; near-zero indicates technicians ignoring the early-warning band.
• Drift trending (as-found delta vs previous as-left) — per-instrument trend; positive slope predicts upcoming OOT; feeds preventive-maintenance scheduling and instrument-retire decisions.
• Reference-traceability completeness — fraction of calibration records with reference cert ID + valid-through + uncertainty all captured; target 100%; any gap is an ISO/IEC 17025 deficiency.
• Vendor-calibration as-found compliance — fraction of vendor-performed calibrations returning with as-found readings captured at the vendor site; target 100%; gap is a contract-management + QA issue.

1. Instrument master carries asFoundTolerance + asLeftTolerance + adjustmentTrigger + calibrationSopId + referenceStandardClass + nextCalibrationDue + calibrationValidUntil — all under change-control with QA approval required for tolerance edits.
2. Calibration record schema enforces required fields: reference identity + cert + uncertainty (hard-block if cert expired or uncertainty missing), per-test-point as-found readings (hard-block on save if any test-point as-found is blank), per-test-point as-left readings, adjustment events captured separately with what + by-whom + authority, two e-sigs (technician + independent reviewer).
3. OOT detection is engine-side, not technician-side: any as-found reading outside the as-found tolerance auto-opens a deviation + auto-queries every dispense_result / batch / WO since previous as-left + pre-populates the back-impact workflow.
4. Instrument status state machine: in-service → calibration-due (soft-warning) → calibration-due-hard-block (grace expired) → out-of-service-pending-investigation (OOT detected) → in-service (only after deviation closure if OOT, or after as-left pass if routine).
5. Calibration kiosk widget runs the technician through the chain in order: pre-flight (cert valid + reference identity) → as-found capture (test-point by test-point, raw reading + tolerance evaluation live) → adjustment capture (if performed) → as-left capture → e-sig → reviewer e-sig.
6. Back-impact workflow drives a per-batch QRM scoring screen for each affected batch + a §211.192 review item written to each affected BMR + a quarantine flag on any high-risk batch.
7. Adjustment-required-threshold triggers an in-line preventive-adjustment prompt + a separate non-OOT event flagged in trending — drift caught before OOT is logged as preventive success, not as a deviation.
8. Vendor-calibration contracts require as-found capture at vendor site + verification at site receipt; vendor calibration records cross-uploaded; missing as-found is auto-flagged as supplier-quality issue + scorecard hit.
9. Calibration PDF generated from the as-found / as-left record + cross-referenced from instrument master + from every dispense_result that consumed the prior as-left; the PDF is rendered through @react-pdf/renderer + stored in the regulated-reports bucket scoped by tenant.
10. Quarterly product review (ICH Q10 §3.2.5) auto-aggregates per-instrument OOT rates + back-impact extents + adjustment trends + tolerance breach frequencies; chronic OOT triggers a capability-ratio + PM-frequency revisit, not a calibration-frequency-only response.