CSV Deviation Report
A CSV deviation report is a controlled record of a failed test step, unexpected result or departure from the validation plan during computerised system validation, with impact assessment and resolution before release.

On this page · 17 sections
- 1What a CSV deviation report is
- 2Common causes
- 3What the report records
- 4Classifying impact
- 5Resolution and retest
- 6Common pitfalls
- 7Inside V5: deviations and validation support
- 8Explain it for your operation
- 9A practical example
- 10Risk-based effort
- 11Supplier defects and known issues
- 12Trending deviations across projects
- 13Summarising in the validation report
- 14A proportionate implementation plan
- 15Internal review questions
- 16Records and evidence to retain
- 17Questions to ask when evaluating software
How does CSV Deviation Report apply to your shop floor?
Pick your industry and scale — Ask V5 rewrites the definition in your context, gives a worked example, and shows what V5 does on day one.
01What a CSV deviation report is
During computerised system validation, testers execute scripts and compare actual results with expected results. When they differ, or when the team departs from the approved plan, the event is captured in a deviation report rather than quietly corrected.
The report keeps the validation honest: reviewers can see what failed, why, and how the team decided it was acceptable to proceed.
02Common causes
- Software defect: the system does not behave as specified.
- Script error: the expected result or steps were wrong.
- Tester error: steps executed incorrectly.
- Environment or data issue: wrong configuration, test data or interface.
- Planned departure: scope or sequence changed from the approved plan.
03What the report records
A useful report identifies the script and step, describes expected and actual results with evidence such as screenshots, classifies the cause, assesses impact on the intended use and on other tests, and records the resolution and any retesting.
It ends with review and approval by the people responsible for the validation and for quality.
04Classifying impact
Teams usually grade deviations by risk to intended use: a defect that could corrupt a record or bypass a signature differs from a cosmetic label error. The grade drives whether release is blocked, whether a workaround is acceptable and how much retesting is needed.
FDA's software assurance guidance encourages focusing effort on features whose failure would affect product quality or patient safety.
05Resolution and retest
Resolutions include fixing the software and retesting, correcting the script with justification, accepting a known limitation with a procedural control, or deferring a feature from scope. Each needs a rationale.
Open deviations should be summarised in the validation report so the release decision is made with full knowledge.
06Common pitfalls
- Rewriting expected results after the fact without a deviation.
- Closing defects as 'tester error' without evidence.
- Retesting the fixed step but not dependent functions.
- Accepting workarounds that are never written into procedures.
07Inside V5: deviations and validation support
Deviation and CAPA workflows in V5 can record a validation finding with attachments, impact assessment, owner, actions and electronic signatures where configured, all with an audit trail.
For Enterprise customers, V5 may include IQ/OQ and validation support. Your organisation remains responsible for its validation approach, performance qualification and the decision to release the system for use.
08Explain it for your operation
A pharmaceutical site validating an electronic batch record will see deviations around signatures, calculations and interfaces. A device maker validating production software focuses on functions that affect device quality. A food plant may use a lighter, risk-based approach for its quality system software.
Agree before testing starts how deviations will be graded, who approves them and what blocks release.
09A practical example
During OQ of a weighing interface, a script expects a tolerance warning at 2% over target. The system warns at 2.5%. The tester records actual results and screenshots and raises a deviation.
Investigation shows the configured tolerance was wrong, not the software. The configuration is corrected under change control, the step and two dependent scripts are rerun, and quality approves closure before the validation report is signed.
10Risk-based effort
Not every deviation needs the same depth. A failure in a function that calculates release results, enforces signatures or protects records deserves full investigation and broad regression testing. A misaligned screen label may need only a correction and a note. The validation plan should define these grades before testing starts.
Risk-based testing guidance from FDA supports concentrating effort where failure would affect product quality, patient safety or record integrity.
11Supplier defects and known issues
Some deviations trace to defects in supplier software. The report should record the supplier's reference, any workaround and the planned fix, and the team should assess whether supplier release notes and known-issue lists reveal related risks.
Supplier documentation can reduce your own testing, but the decision on fitness for intended use remains with the regulated company.
12Trending deviations across projects
Looking across validation projects shows patterns: script errors may point to weak requirements, repeated environment issues may point to poor test-system control. Periodic review of deviation categories helps improve the next project, not just close the current one.
13Summarising in the validation report
The final validation report should list every deviation, its classification, resolution and status. Reviewers rely on this summary to judge whether the system is fit for release, so it should be complete and consistent with the individual reports.
14A proportionate implementation plan
- Agree the scope, the decision the records support and who owns it.
- Map the current records and find where evidence is missing or duplicated.
- Write or update the procedure, with clear responsibilities and escalation.
- Pilot on one area with real data and review what was hard to use.
- Train the people involved and confirm understanding.
- Review results after a defined period and adjust under change control.
15Internal review questions
- Can we show the procedure that defines this activity and who approved it?
- Can we reconstruct a recent example from the records alone?
- Are exceptions recorded, assessed and closed with a rationale?
- Who reviews the data, how often and with what evidence of review?
- What changed since the last review, and was the change controlled?
16Records and evidence to retain
Keep the approved procedure, the source records, exceptions and their assessments, review and approval evidence, and any resulting actions, with enough context for someone unfamiliar with the event to understand what happened and why.
17Questions to ask when evaluating software
- Can users trace every summary value back to its source record and version?
- How are roles, permissions, signatures and corrections recorded?
- Can the workflow reflect our approved procedure without inventing a universal rule?
- What happens during an outage, and how are later entries distinguished?
- Can we test the highest-risk exception with realistic data before committing?
- Which decisions remain explicitly assigned to an authorised person?
Frequently asked questions
Q.What is a CSV deviation report?+
It is a controlled record of a failed test step, unexpected result or plan departure during computerised system validation, with impact assessment and resolution.
Q.Can a system be released with open deviations?+
Only if they are assessed, justified and controlled, and the release decision is documented by the responsible people.
Q.Who approves CSV deviations?+
Typically the validation owner and the quality unit, as defined in the validation plan.
Q.Can software make the final decision?+
No. Software can organise records, calculations and workflow, but authorised people remain responsible for quality and disposition decisions.
Q.How should a site start?+
Choose one bounded process, define the decision and evidence, test the method with real records, correct gaps, then expand under change control.
Q.What evidence should be retained?+
Keep the approved method, source records, exceptions, review, decision, approval and any resulting action with enough context to reconstruct the work.
Q.Is a CSV deviation the same as a production deviation?+
No. A CSV deviation concerns validation testing of a computerised system. A production deviation concerns a departure during manufacturing. Both follow a controlled process, but they feed different records and decisions.
Q.Should script errors be raised as deviations?+
Yes. Correcting a wrong expected result without a deviation hides the change; the report records the error, the justification and the corrected script.
Primary sources
Want to see how CSV Deviation Report could fit into your own records and workflows? Explore the related V5 pages or talk to our team about what applies to your operation.
