GMP documentation is meant to demonstrate that a system, a process or a product was planned, implemented and tested in a controlled way. In practice, this evidence rarely fails because of a single major omission. It fails because of many small inconsistencies that add up across project phases and documents. The errors below come up again and again in qualification and engineering projects, whether for cleanrooms, HVAC systems and compressed air systems or for laboratory equipment. For each point, you will find why it is a problem and how to avoid it.
Errors in the requirements
Requirements that cannot be tested
Phrases such as “adequate air change rate” or “user-friendly operation” cannot be tested. By the OQ at the latest, there is no objective criterion, and the test becomes a matter of judgement.
How to avoid it: Every requirement in the user requirement specification gets a unique ID, a measurable target value or a clearly observable behaviour, and a note on how it is to be verified. A good check question is: could someone who was not involved decide from this requirement whether it has been met?
Acceptance criteria missing or adjusted afterwards
If the test plan has no acceptance criteria, it looks as though the result was only assessed once it was available. It is even more critical when criteria are quietly adjusted to fit the result after the measurement.
How to avoid it: Acceptance criteria are defined before execution and approved together with the test plan. If a criterion has to change significantly during execution, Annex 15 (2.7) expects this to be documented as a deviation and scientifically justified, not handled by correcting the completed protocol. Later changes to acceptance criteria should also be scientifically justified (2.9).
Errors in structure and linking
Missing or broken traceability
If you cannot trace which requirement is covered by which test, it remains unclear whether all critical points were tested. Typical examples are requirements without a test, tests without a corresponding requirement, or IDs that differ between the URS and the protocol.
How to avoid it: A maintained traceability matrix links requirement, risk, specification and test case. It is not created at the end of the project. It grows with the project and is fully reconciled before the final report.
Inconsistent versions and document numbers
A protocol refers to version 1.0 of the URS, but version 2.0 has already been approved. Or the same document number appears with different titles. In an audit, this immediately raises the question of which version was actually tested against.
How to avoid it: Keep a central document register with number, title, version and status. In practice, it pays to enter the actual document ID and revision as shown on the document itself, rather than an internal project shorthand. References always state the number and the version. Before each approval, check that the referenced documents are valid in the version cited. In engineering projects, this includes reconciling drawing revisions between trades, because HVAC, electrical and architectural drawings often evolve at different speeds.
References that lead nowhere
Cross-references to attachments, drawings or calibration certificates that are not attached or cannot be found make the evidence incomplete, even if the work itself was correct.
How to avoid it: During document review, actually resolve every reference. Attachments are numbered uniquely and listed in the main document with their page count.
Copies from templates without adaptation
Templates save time, but they carry the risk that test steps, equipment designations or limits from another project are left in. A protocol that names the wrong room number or components that do not exist damages the credibility of the entire document set.
How to avoid it: Templates contain clearly marked placeholders. Test steps that do not apply are not deleted but marked “not applicable” with a justification. The technical review is carried out by someone who knows the specific system.
Errors in execution and evidence
Missing raw data
A protocol with the entry “differential pressure OK” is not evidence without the measured value, the measuring instrument and the time. The same applies to printouts from measuring instruments, data exports from the building management system or particle counter records that were not retained.
How to avoid it: Measured values are recorded with the unit, the measuring instrument and its calibration status, the date and the person who carried out the test. Original printouts and electronic raw data are assigned to the protocol and stored so that they remain readable throughout the retention period.
If an external laboratory or a specialist measurement contractor supplies a report, its conclusion does not replace your own assessment against the approved acceptance criteria. It has proven useful to keep the measuring point IDs identical in the raw data, the report and the traceability matrix, so that every value can be assigned directly.
Deviations not documented or not assessed
Deviations are normal in qualification work. They become a problem when they are only resolved verbally, written over in the protocol, or recorded but never assessed for their impact on product quality and qualification status.
How to avoid it: Every deviation gets a number, a description, a root cause review, an assessment and an action. The final report lists all deviations with their status. Open items are not concealed but stated with a justification and a deadline.
In larger projects, we also keep a central deviation list. For each deviation, it shows the phase in which it was found, its classification (critical, major, minor), the assessment and the phase by which it must be closed. The list provides an overview. It does not replace the formal handling of deviations and CAPA in the quality management system.
Incomplete signatures and approvals
Missing reviewer or approver signatures, undated signatures, or an approval given after execution has already started are classic audit findings. They call into question whether the intended control actually took place.
How to avoid it: Roles and responsibilities are defined in the qualification plan. Before execution starts, check that the test plan has been approved. For electronic signatures, section 14 of Annex 11 of the EU GMP Guide applies, and for the US market 21 CFR Part 11.
Errors in the lifecycle
Changes without change control
After qualification, a setpoint is adjusted, a filter is sized differently or software is updated, without a formal change assessment. The documented qualification status then no longer matches the actual state of the system.
How to avoid it: Every change to qualified systems goes through the change control procedure. That is where you assess which documents are affected and whether requalification or a partial test is needed. The traceability matrix and the document register are updated accordingly.
Breaches of good documentation practice
Many findings concern the form of the record rather than its technical quality. Chapter 4 of the EU GMP Guide sets out specific rules for this. Handwritten entries should be clear, legible and indelible (4.7) and made at the time each action is taken (4.8). Alterations are signed and dated, the original information remains readable, and where appropriate the reason is recorded (4.9). The integrity of records must be ensured throughout the retention period (4.10), and the expected documentation includes signature lists (4.29).
ALCOA+ is a widely used mnemonic for data integrity. It summarises the properties that data and records should have:
| Principle | Typical breach | Practice |
|---|---|---|
| Attributable | Initials without a signature list | Keep a signature list for the project |
| Legible | Overwriting, correction fluid, pencil | Single strike-through, correction with date and initials, reason where not obvious |
| Contemporaneous | Filling in later from memory | Enter values at the time of measurement |
| Original | Only transcribed values, original discarded | Keep and reference the original record |
| Accurate | Calculation or transcription errors | Show calculations traceably, four-eyes check |
| Complete, consistent, enduring, available | Empty fields, loose sheets, illegible thermal printouts | Strike out empty fields, number pages, copy and verify thermal printouts |
These rules are simple, but everyone involved must know them, including service providers and fitters who fill in protocols on site. A short briefing before execution starts prevents a lot of rework.
Conclusion: consistency beats volume
Audit-ready documentation is not defined by its page count. It is defined by requirements, tests, results and deviations fitting together completely and without contradictions. If you plan for testability, traceability and documentation discipline from the start, you avoid time-consuming rework before an inspection. You will find more advice on preparation in the article on audit preparation for technical GMP projects.
You can get support with setting up and reviewing your documents through our GMP documentation service.