GxP & Validation · Deep Dive
Software validation under GxP: what teams have to evidence
Validation is not a stack of documents but documented proof of fitness for a defined purpose. This deep dive walks through the process step by step: what is planned, tested, documented and released – and what happens after release.
- Reading time
- approx. 5 minutes
- Last reviewed
- August 2026
- For
- QA, validation, system owners
“We have to validate the system” is a sentence that triggers panic or a shrug in projects – rarely clarity. Yet the process is easy to describe. This deep dive walks through it step by step: what is planned, tested, documented and released, and what happens afterwards.
1. What exactly is being evidenced
Validation demonstrates that a system is fit for its defined intended use – documented, traceable and in the actual operating environment. Three points follow directly:
- Without a defined intended use, validation cannot be performed, because the reference point is missing.
- Validation happens at the operator, not at the vendor. A product can provide prerequisites, nothing more.
- The evidence applies to one version. After changes it has to be maintained or renewed.
2. The process at a glance
- 1 Plan. Intended use, risk assessment, scope, responsibilities, acceptance criteria.
- 2 Specify. Requirements with identifiers, GxP relevance marked, design documented traceably.
- 3 Test. Qualification of installation, function and performance in the operating context.
- 4 Assess and release. Handle deviations, write the report, release by defined roles.
- 5 Maintain. Change control, periodic review, decommissioning with data migration.
3. The validation plan
The plan is written before testing and defines what counts as evidence. It should contain: scope and boundaries, risk assessment, planned qualification steps, responsibilities, how deviations are handled, supplier assessment, retention and the criteria for release.
A common mistake is too broad a scope. Including everything looks thorough but makes validation expensive and sluggish. A clear boundary – what is not in scope – is as important as the inclusion list.
4. Requirements and specification
Requirements are the reference point for everything else. Marking them by GxP relevance helps: not every requirement is regulatorily significant, and that distinction saves considerable effort. Every requirement needs a stable identifier, a testable wording and a link to design, test and release – see traceability.
5. Qualification: IQ, OQ, PQ
| Step | Question | Typical evidence |
|---|---|---|
| IQ | Is it installed correctly? | versions, configuration, environment, checksums |
| OQ | Does it work as specified? | test cases per requirement, including negative tests |
| PQ | Does it work in real use? | testing with realistic data, roles and workflows |
Automated tests cover the OQ area particularly well – provided environment, test data and logging are controlled. PQ usually remains partly manual, because that is where the interplay with people and processes is tested.
6. Deviations and assessment
Deviations are normal. What matters is that they are documented, assessed and decided on – and do not vanish through a quiet adjustment of the test case. A sound approach covers:
- a description of the deviation and the affected test case,
- an assessment of the impact on product quality, patient safety and data integrity,
- a decision: fix and retest, accept with justification, or restrict the intended use,
- approval of that decision by the responsible role.
In short
A documented and assessed deviation is a sign of a working process. A validation report without a single deviation raises more questions than it answers.
7. Report and release
The validation report summarises what was tested, with what result, which deviations occurred and how they were assessed. It ends with a clear statement on fitness for the defined intended use – with restrictions where applicable. Release is given by the roles named in the plan, in particular the quality unit acting independently of development.
8. Maintaining the validated state
After release the longer part begins. It includes:
- Change control for every change to system, configuration or environment,
- periodic review at defined intervals – do the processes still run as described, is the evidence still there?
- handling defects and security updates, including assessing their relevance to validation,
- decommissioning with a defined data migration and readability across the retention period.
9. Keeping the effort realistic
Three levers lower the effort without lowering the quality of the evidence:
- Use the risk basis consistently. Do not test everything to the same depth.
- Credit the supplier's work instead of re-evidencing standard functions.
- Generate evidence automatically instead of writing it – see evidence by design.
Checklist: a validation project
- Intended use and scope are bounded – including what is out of scope.
- The validation plan is approved before testing starts.
- Requirements are marked, testable and linked.
- Test evidence is attributed to the exact version and retained.
- Deviations are documented and assessed, not defined away.
- Release is given by defined, independent roles.
- Change control and periodic review are in place.
Conclusion
Validation is not a documentation project but a testing project with retained results. Define the intended use sharply, plan on a risk basis, credit the supplier's work and generate evidence automatically, and you reach a far more solid result with far less effort.
The expensive part is rarely the testing itself. It gets expensive when requirements are vague, evidence is missing, or the validated state is not maintained after the first update.
Sources & further standards
-
EudraLex Volume 4 – EU GMP guidelines
In particular Annex 11 “Computerised Systems” and Annex 15 “Qualification and Validation”; European Commission. health.ec.europa.eu -
21 CFR Part 11 – Electronic Records; Electronic Signatures
US Food and Drug Administration, version as published in the eCFR. www.ecfr.gov -
ISPE GAMP 5 (2nd edition)
Industry guide for risk-based validation of computerised systems, including software categories; available from ISPE. -
PIC/S PI 011 – Good Practices for Computerised Systems in Regulated “GxP” Environments
Guidance from the Pharmaceutical Inspection Co-operation Scheme. picscheme.org
This article explains terms and relationships. It is neither a regulatory assessment nor legal advice. Which requirements apply to your system, and which evidence is needed, depends on the intended use, the applicable regulations and your quality management system.
Further reading
Related deep dives.
GxP & Validation
GxP software development: requirements, validation and traceability
Intended use, risk-based validation, audit trails, change control.
GxP & Validation
GAMP 5 in modern software projects
The guide behind the risk-based approach.
Platform
How CodamAI implements these steps
MCP integration, explicit backend models, the Hub, OpenAPI and your own CI/CD.
Does CodamAI fit your engineering setup?
The most honest way to find out is a technical conversation: about your stack, your delivery and the requirements you have to evidence.