This is a ready-to-use working paper for capturing audit evidence on the floor. Replace every <<FILL: ...>> placeholder with your own specifics, set your document numbers and dates, and route the completed form through your normal records retention. A worked filled specimen follows the template. Verify each cited regulation against the current published source before you rely on it. This template is an educational reference for you to adapt to your own quality system, products, and regulatory context; it is not legal, regulatory, or professional advice.
The working paper is the raw evidence behind an audit report. Everything in the report should be traceable to a row on this form, captured at the moment it was seen. Auditors who plan to write it all up that evening produce reports full of “documentation was inconsistent” instead of “logbook CL-204, entries 12 to 18 March 2026, 9 of 14 missing the second signature required by SOP-CLN-007 rev 4 step 6.2”. The difference is not memory. It is whether you wrote down the identifier while you were standing there.
Form control
| Field | Entry |
|---|---|
| Form number | <<FILL: e.g. FRM-QA-052>> |
| Version | <<FILL>> |
| Audit reference | <<FILL: audit ID>> |
| Supplier and site | <<FILL>> |
| Audit dates | <<FILL>> |
| Auditor (this working paper) | <<FILL: name and initials used below>> |
| Sheet | <<FILL: n>> of <<FILL: n>> |
| Confidentiality | <<FILL: e.g. Confidential, internal audit working paper, not for distribution to the supplier>> |
Each auditor keeps their own sheets and initials their own rows. Do not pool observations onto one sheet written by whoever had the pen; attribution matters if a finding is later challenged.
The evidence log
One row per observation. A row is created whether or not the observation becomes a finding, because the observations that turned out to be fine are what show the audit had coverage.
| Ref | Date | Time | Area | Document or record ID and version | Specific entry, lot, instrument or person | Factual observation | Possible requirement breached (external) | Possible requirement breached (supplier’s own SOP) | Auditor initials | Raised with host on the floor (Y/N, time) | Disposition |
|---|---|---|---|---|---|---|---|---|---|---|---|
<<FILL: 001>> | <<FILL>> | <<FILL>> | <<FILL>> | <<FILL>> | <<FILL>> | <<FILL>> | <<FILL>> | <<FILL>> | <<FILL>> | <<FILL>> | <<FILL>> |
<<FILL: 002>> | |||||||||||
<<FILL: 003>> | |||||||||||
<<FILL: 004>> | |||||||||||
<<FILL: 005>> |
Disposition values
| Value | Meaning |
|---|---|
| OBS | Observation only. Evidence supported the requirement, or the point is context. No further action. |
| PF | Potential finding. Carried to the team caucus for grading. |
| PF-CONF | Potential finding, confirmed by additional evidence. Spread established. |
| WD | Withdrawn. The supplier produced evidence that resolved it. Record what that evidence was. |
| OPEN | Not yet resolved at the time of writing. Must be closed out before the closing meeting or reported as an unresolved item. |
Nothing may enter the audit report unless it appears on a working paper row with a disposition of PF or PF-CONF. That single rule kills the two worst report defects at once: findings invented at the writing desk, and findings that first appear after the closing meeting.
Field definitions
| Field | Format | Required | Who completes it | When |
|---|---|---|---|---|
| Ref | Sequential number, unique within the sheet, e.g. 001 | Yes | The auditor who observed it | At the moment of observation |
| Date | DD MMM YYYY | Yes | Observing auditor | At the moment of observation |
| Time | 24-hour clock, HH:MM | Yes | Observing auditor | At the moment of observation |
| Area | Free text, the physical or system area, e.g. warehouse cold store, QC HPLC lab, change control review | Yes | Observing auditor | At the moment of observation |
| Document or record ID and version | Exact identifier as printed on the record, plus revision or version and effective date where shown | Yes for any row that may become a finding | Observing auditor | At the moment of observation. Do not write “the cleaning SOP” |
| Specific entry, lot, instrument or person | The row, page, entry date, lot number, equipment ID, sequence name, or the role of the person interviewed | Yes for any row that may become a finding | Observing auditor | At the moment of observation |
| Factual observation | What was seen or said, in the past tense, with counts. Affected count and examined count both. No adjectives, no inference, no proposed fix | Yes | Observing auditor | At the moment of observation |
| Possible requirement breached (external) | Regulation, guideline, quality agreement clause, or specification, cited to part and section | Where a potential finding is suspected | Observing auditor, may be refined at caucus | Same day |
| Possible requirement breached (supplier’s own SOP) | The supplier’s procedure number, revision and step | Where a potential finding is suspected | Observing auditor | Same day. Ask for the SOP while you are still in the area |
| Auditor initials | Two or three letters, matching the form control header | Yes | Observing auditor | At the moment of observation |
| Raised with host on the floor | Y or N, with the time and the name of the person told | Yes | Observing auditor | At the time of raising |
| Disposition | One of OBS, PF, PF-CONF, WD, OPEN | Yes | Observing auditor, confirmed at team caucus | Set on observation, updated at caucus |
Rules for completing this form
- Capture at the moment of observation, with full identifiers. Not at the coffee break, not that evening. A record identifier you did not write down is a record you cannot cite, and the supplier’s help in reconstructing it later is not evidence.
- Write what you saw, not what you concluded. “Nine of fourteen entries have no second signature” is an observation. “Cleaning verification is not being performed” is a conclusion, and it may be wrong: the verification might happen and be recorded elsewhere. Separate the two so that if your inference collapses, the fact survives.
- Always record the denominator. Three deviations without impact assessment means something different when three of three were examined than when three of forty were. The denominator is what establishes spread, and spread is what drives the grade.
- Cite both bars. The external requirement and the supplier’s own procedure. Get the supplier’s SOP number and revision while you are still standing in the area, because retrieving it later takes an hour and sometimes produces a different revision.
- Raise it on the floor and record that you did. The “raised with host” column is what lets the lead auditor state at the closing meeting that nothing is a surprise, and lets the report state the same thing truthfully.
- Do not write the fix. Root cause and corrective action belong to the supplier. An auditor who writes “they should add a second verification step” has stopped auditing and started consulting, and has weakened both the finding and the supplier’s ownership of it.
- Do not erase or overwrite. If an observation is later shown to be wrong, add a new row or annotate with a single-line strikethrough, your initials and the date, and state what evidence changed it. Working papers follow the same documentation practice as any other GxP record.
- Photography and copying follow the site policy. Where photography is prohibited, transcribe the entry verbatim with its identifiers and note in the row that the entry was transcribed. Where the supplier provides a copy, record the copy reference in the row.
- One auditor, one sheet, own initials. Attribution survives challenge; pooled notes do not.
- Close every OPEN row before the closing meeting. An unresolved OPEN either becomes a PF, becomes a WD with the resolving evidence recorded, or is reported as an item the audit could not conclude on.
Retention of working papers
| Field | Entry |
|---|---|
| Retention period | <<FILL: period, aligned to the audit report retention in your records schedule>> |
| Storage location | <<FILL: controlled location or system>> |
| Access | <<FILL: e.g. quality assurance and the audit team only>> |
| Distribution to the supplier | <<FILL: normally none. Working papers support the report; the report is the deliverable>> |
| Handling of supplier-provided copies attached to these papers | <<FILL: retained under the confidentiality agreement, marked as supplier confidential>> |
Working papers are retained with the audit report and for the same period. They are the evidence that the audit examined what it claims to have examined. An audit file containing a report and no working papers cannot demonstrate that anything was actually seen, and an audit file whose working papers were destroyed after the report was issued is worse than one that never had them.
Sheet certification
| Field | Entry |
|---|---|
| Rows on this sheet | <<FILL: count>> |
| Rows with disposition PF or PF-CONF | <<FILL: refs>> |
| Rows with disposition OPEN at end of audit | <<FILL: refs, and how each was resolved>> |
| Auditor name and signature | <<FILL>> |
| Date completed | <<FILL>> |
| Reviewed by lead auditor | <<FILL: name, signature, date>> |
Filled specimen
A completed sheet from an illustrative on-site audit of a contract sterile fill-finish supplier. Company, records, people and dates are invented for teaching purposes.
Form control: FRM-QA-052 rev 3. Audit SUP-AUD-2026-018, Northvale Sterile Services Ltd, Cork. Audit dates 19 to 20 May 2026. Auditor: P. Raghavan, initials PR. Sheet 2 of 3.
| Ref | Date | Time | Area | Record ID and version | Specific entry | Factual observation | External requirement | Supplier SOP | Init | Raised on floor | Disp |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 011 | 20 May 2026 | 09:20 | QC lab, chromatography data system | CDS configuration screen, software v8.4, viewed live with administrator T. Nolan | Project NORTHVALE_QC, audit trail settings tab | Audit trail active at project level. “Disable audit trail” option greyed out for Analyst and Reviewer roles; available to System Administrator role, held by two named IT staff who do not generate data. Demonstrated on a test record that old and new values are captured. | EU GMP Annex 11 section 9; 21 CFR Part 11 | SOP-IT-009 rev 2 s5.1 | PR | N | OBS |
| 012 | 20 May 2026 | 10:05 | QC lab | Sequence NV-HPLC03-260412-A, CoA NS-26-0031 | Assay result 99.2 percent, reported 13 Apr 2026 | Backward trace completed from CoA NS-26-0031 to the sequence, the raw data file, instrument HPLC-03 and analyst L. Moran. Result on CoA matches the processed result in the system. Audit trail for the sequence shows 6 injections, no re-integration, no aborted run, no renamed file. | 21 CFR 211.194 | SOP-QC-022 rev 7 | PR | N | OBS |
| 013 | 20 May 2026 | 10:48 | QC lab | Audit trail review forms ATR-2026-041, -052, -061 | Reviews for the sequences supporting lots NS-26-0031, NS-26-0072, NS-26-0114 | All three forms signed and dated within the required window. Each records only the phrase “audit trail reviewed, no issues”. No entry count, no filter or query named, no statement of what categories were examined. 3 of 3 reviews examined. Reviewer A. Cronin, asked at 10:52, described opening the sequence trail and scanning for re-integrations and aborted runs; that method is not documented on the form or in the SOP. | EU GMP Annex 11 section 9 | SOP-QC-030 rev 4 s7.2 requires documented review of the audit trail for every release sequence before the result is reported | PR | Y, 11:05, told QC manager S. Devlin | PF |
| 014 | 20 May 2026 | 11:20 | QC lab | User access list, CDS, exported 20 May 2026 | 34 active accounts | 34 active accounts. Two accounts named QCLAB1 and QCLAB2 are not attributable to a named individual. Asked S. Devlin: stated they are legacy accounts used for instrument control only and not for data acquisition. Requested the audit trail for both accounts for the last 90 days. | 21 CFR Part 11; EU GMP Annex 11 | SOP-IT-009 rev 2 s6.3 | PR | Y, 11:22, S. Devlin | OPEN |
| 015 | 20 May 2026 | 14:10 | QC lab | Audit trail export for QCLAB1 and QCLAB2, 19 Feb to 20 May 2026, provided by T. Nolan as copy REF-A7 | All entries | Export reviewed. QCLAB2 shows 4 sample acquisition events on 03 and 04 Mar 2026, sequences NV-HPLC01-260303-B and -260304-A, both supporting release testing of lot NS-26-0044, a lot supplied to us. Acquisition under a non-attributable shared account contradicts the account’s stated instrument-control-only use. 2 of 2 sequences in the period under this account were acquisitions, not instrument control. | 21 CFR Part 11 requires records to be attributable to individuals; EU GMP Annex 11 section 12 on access control | SOP-IT-009 rev 2 s6.3 prohibits shared accounts for GxP data entry | PR | Y, 14:30, S. Devlin and QA head M. Byrne | PF-CONF |
| 016 | 20 May 2026 | 15:05 | QC lab, reference standards | Standard register REG-STD-04, entry 118 | Working standard WS-ASSAY-11, expiry 30 Jun 2026 | Standard used in the traced result at ref 012 is in date, has a valid certificate traceable to the compendial standard, and storage conditions match the certificate. Retention sample present. | 21 CFR 211.194 | SOP-QC-014 rev 3 | PR | N | OBS |
| 017 | 20 May 2026 | 15:40 | QC lab | OOS log 2025 to 2026, and investigation OOS-25-031 | Investigation for a dissolution failure on lot NS-25-0918, closed 04 Dec 2025 | Phase 1 laboratory investigation concluded “analyst error, dilution step” and invalidated the original result. Investigation record contains no evidence supporting the dilution error: no interview record, no re-check of the original preparation, no repeat under the original conditions. Result was invalidated and a retest passed. 1 of 5 OOS investigations reviewed shows invalidation; that one lacks assignable-cause evidence. | 21 CFR 211.192 requires a thorough investigation with conclusions and follow-up | SOP-QC-025 rev 5 s8.3 requires documented objective evidence of a laboratory error before a result may be invalidated | PR | Y, 16:00, M. Byrne | PF |
Sheet certification: 7 rows. PF or PF-CONF: 013, 015, 017. OPEN at end of audit: none; ref 014 was resolved by ref 015. Completed by P. Raghavan, 20 May 2026. Reviewed by lead auditor K. Ferreira, 20 May 2026.
Three things in this sheet are worth copying. First, rows 011, 012 and 016 are Pass observations, and they belong on the paper. They are what proves the audit went into the system rather than around it, and they are the reason nobody can later claim the data integrity review was superficial.
Second, look at the sequence 014 to 015. Ref 014 was a soft observation with an explanation from the site that sounded reasonable. The auditor did not accept the explanation; he requested the evidence that would test it and left the row OPEN. The evidence came back at 14:10 and turned an “explained” observation into a confirmed finding on a lot we had actually received. That is what OPEN is for.
Third, notice how narrow the observations are. Ref 017 says the investigation record contains no supporting evidence. It does not say the analyst did not make an error, and it does not say the result was falsified. If the supplier later produces an interview record that was misfiled, the finding weakens but the auditor is not exposed, because the auditor only ever claimed what was in the record at the time it was read.
Common inspection findings this form prevents
- Audit findings written as opinions with no record identifier, version or date, so the supplier cannot act on them and a challenge collapses them.
- Findings that appear for the first time in the report, never raised with the supplier during the audit.
- Audit files containing a report but no underlying evidence, so no one can demonstrate what was examined.
- Working papers destroyed or discarded after the report was issued, leaving the report unsupported.
- Findings with no denominator, so spread and therefore grade cannot be justified or defended.
- Findings citing a regulation but never the supplier’s own procedure, allowing the supplier to argue the bar was invented.
- Observations rewritten from memory at the end of the day, with identifiers that turn out not to exist.
- An audit that recorded only problems, with no evidence of the areas examined and found acceptable, so coverage cannot be shown.
- A potential finding explained away on the floor and never tested against evidence, because there was no mechanism to leave it open.
How to adapt this form
- Set the form number and retention to your own records schedule, and confirm that working papers are retained for the same period as the audit report.
- If your auditors work on tablets, keep the same columns but make the record identifier and denominator fields mandatory so a row cannot be saved without them.
- Add a column for photograph or attachment reference if your sites permit photography; keep the transcription rule for sites that do not.
- For remote or documentary audits, replace “raised with host on the floor” with “raised on the call, with date and time”, and add a column for the file name and hash or version of any electronic evidence received.
- Add a column for quality agreement clause if most of your audits test contractual commitments; it saves cross-referencing later.
- Keep the disposition values short and defined. Auditors invent their own codes when the list does not fit, and then nobody can total the potential findings.
- Carry every PF and PF-CONF row into Report: On-Site GMP Supplier Audit and grade it with SOP: GxP Audit Finding Classification and Grading. Use this form alongside Checklist: On-Site GMP Supplier Audit Fieldwork, which drives coverage while this form captures the evidence. The method behind both is described in how to conduct a supplier or GMP audit.