Self-Assessment vs Quality Audit: What Each Can Establish
Self-assessment vs quality audit comes down to evidence and independence: one records what a team reports about its own practices, the other examines evidence against criteria. Knowing which one you are holding decides what you may conclude from it.
Written by CSV/CSA Maturity Editorial TeamPublished 9 min read
On this page8 sections
A self-assessment and a quality audit establish different things. A self-assessment records what a team reports about its own practices, so it can show where the team believes it stands and what it cannot yet demonstrate. A quality audit examines evidence against defined criteria with a stated degree of independence, so it can show whether the sampled practice met those criteria at the time. The first gives you prompts for planning. The second gives you findings that your quality system has to handle.
This comparison is for quality leaders and system owners reading the result of a maturity self-assessment for computer system validation (CSV) or computer software assurance (CSA), including the one on this site. CSV here does not mean comma-separated values. The article does not define an audit standard. What counts as an audit in your organization, who may perform one and what follows from it are set by your own quality system and by the requirements that apply to you.
Three activities that are easy to confuse
Between a self-assessment and a formal audit sits a third activity that teams sometimes perform without naming it: an internal evidence review. A colleague pulls a few records and checks whether they support what the team said about itself. It is less than an audit and more than an opinion, and it is often a sensible next step after a self-assessment.
The table compares the three on the points that decide what you may conclude.
| Criterion | Self-assessment | Internal evidence review | Formal audit process |
|---|---|---|---|
| Purpose | Describe current practice as the team sees it and decide where to look first | Check whether selected answers hold up against actual records | Determine whether practice conforms to defined criteria |
| Scope | Chosen by the respondent: one team, one site or a system portfolio | A few records chosen to test specific answers | Set in an audit plan or program under the quality system |
| Evidence depth | None is examined. Answers rest on recall or on records the team chose to consult | Sample records are read by a colleague | Records, interviews and observation, as far as the audit scope allows |
| Independence | None. The team describes itself | Partial. The reviewer is internal and may know the work | Set by the audit procedure. The auditor may be internal or external |
| Sampling | One answer per question for the whole scope | Small and purposive, aimed at doubtful answers | Planned and recorded, so a reader can see what was and was not examined |
| Output | A self-reported profile, with scores or possible ranges and review prompts | Notes on what was found, revised answers and open questions | Findings or observations, recorded and classified under the audit procedure |
| Follow-through | Voluntary until you enter an action in your own system | Actions agreed with the owner, escalated if something serious appears | Managed through the quality system, for example by corrective action and follow-up |
| Can establish | What the team reports, where knowledge is missing and where to look first | Whether specific answers are supported by the records sampled | Whether the sampled practice met the criteria at the time of the audit |
| Cannot establish | That any practice is in place, effective or compliant | Conformance across the whole scope | Anything outside its scope, its sample or its date |
"Formal" does not mean "external". Audits may be performed by an organization's own audit function, by a customer or partner, or by another party. They differ in mandate, criteria and consequences, and an inspection by an authority is a separate activity again. This article does not rank them.
A planning prompt is not an audit finding
The assessment on this site produces prompts. A prompt is generated by a published rule from the answers you selected. For example, three of its 18 questions carry a priority review prompt: answer one of them with the lowest practice description and the profile asks you to review that practice independently of the average. No evidence sits behind the prompt, nobody classified it, and nothing obliges you to act on it.
An audit finding is a different object. It states what was observed, against which criterion, on the basis of which evidence, and it is recorded under a procedure that usually requires a response. That difference has two practical consequences:
- Do not copy a prompt into an audit report or a quality record as though it were a finding. It may become the reason someone looks, and what they find is the finding.
- Do not cite a self-reported profile as evidence that a practice conforms. The profile itself states that it is "not an audit, certification, regulatory determination or confirmation of compliance."
One exception runs the other way. If the team discovers a real problem while gathering evidence for an answer, such as records that cannot be found, that discovery belongs in your established quality process. It does not wait for the self-assessment to be finished.
Illustrative example: a high score built on unchecked answers
Synthetic sample. The team, answers and numbers below were invented to show how the published scoring model (model version csa-maturity-2.0.0) behaves. They do not describe any organization, an industry average or a target.
A team lead completes the assessment alone, in one sitting, from memory. She selects the highest practice description for 13 of the 18 questions and the second-highest for the other five (Q04, Q10, Q12, Q13 and Q17). The profile reads 91.3 out of 100, in the band the rubric calls Measured, with 18 of 18 numeric answers and no review prompts. With nothing low or unknown, the one action it suggests is to sample a recent change or approval record and confirm that the described practices are applied.
A colleague then asks to see the evidence behind three of the top answers: the evidence version behind an assurance decision (Q09), the controls that protect assurance records (Q15) and a retrieval exercise for an evidence package (Q16). Each of those answers describes a check the team performs. Nobody can produce a record of the check that afternoon. The accurate answer to all three is "I do not know yet".
| Measure | First pass, from memory | After three answers become unknown |
|---|---|---|
| Numeric answers | 18 of 18 | 15 of 18 |
| Numeric coverage | 100.0% | 83.3% |
| Overall result | 91.3, Measured | No score or band. Possible range 74.6 to 91.3 |
| Traceability | 88.9 | Possible range 55.5 to 88.9 |
| Approvals and record controls | 100.0 | Possible range 50.0 to 100.0 |
| Evidence and review readiness | 88.9 | Possible range 55.5 to 88.9 |
| Review prompts | None | Verify the evidence for Q15 |
Read the two columns together:
- Nothing about the team's practice changed between them. Only the claim about what is known changed.
- The top of the possible range, 91.3, is the first score. The lower end of a range substitutes 0 for each unknown answer and the upper end substitutes 3. The first pass was the most favorable reading of three answers nobody had checked.
- The second profile has no overall score and no band. A partial profile never shows one.
- Q15 is one of the three questions that carry a priority review prompt. Answered with the lowest description, it asks for the practice to be reviewed. Answered as unknown, as here, it asks the owner to verify the evidence. Neither changes the arithmetic.
- The possible range is arithmetic. It is not a confidence interval and not a forecast of the score the team will have after checking.
- The second profile is still self-reported. The other 15 answers are exactly as unverified as these three were.
The team lead was confident both times, so confidence was never the missing ingredient. What was missing was a record that someone else could look at. A self-assessment cannot go and look. An evidence review or an audit can. Our tutorial on reading a profile's area scores, numeric coverage and prompts works through the model's published examples in full.
Maturity assessment vs compliance assessment
A maturity assessment asks how established and how reviewed a practice is, on a rubric. This site's rubric has four bands: Ad hoc (0.0 to below 25.0), Developing (25.0 to below 50.0), Defined (50.0 to below 75.0) and Measured (75.0 to 100.0). These are names in a self-report rubric. "Measured" does not mean that anyone inspected the evidence or that controls are effective.
A compliance assessment asks a different question: are the specific requirements that apply to this organization met. Answering it takes three things a maturity questionnaire does not have:
- A determination of which requirements apply, made by people qualified to make it.
- Evidence examined against each requirement.
- An assessor with the competence and the mandate to judge the result.
The assessment says as much in its own wording. The note on its question about relevance to regulated work reads: "This question does not decide which laws apply."
A high band therefore does not show that a given requirement is met, because the rubric does not test requirements. A low band is not a finding of non-compliance, for the same reason. A score of 100.0 is the top of the rubric and not a perfect compliance result.
When a formal review is needed
A self-assessment is enough when the question is "where should we look first". It is not enough when:
- someone outside the team has to rely on the conclusion, such as management, a customer or an authority;
- your quality system, or a requirement that applies to you, calls for an audit or a defined evaluation;
- the self-assessment has surfaced a possible failure of a control rather than a gap in knowledge;
- a decision with consequences for product or patients depends on the answer;
- unknown answers remain unknown after an internal evidence review.
One example of a defined evaluation, from the EU GMP context for medicinal products: Annex 11 [S36] says computerized systems should be periodically evaluated (clause 11). As of November 12, 2026, the EudraLex Volume 4 index [S04] lists Annex 11 as the January 2011 revision. In our view, a self-reported profile is not that evaluation, and completing one does not discharge it. Whether Annex 11 or any other text applies to your systems is for your quality and regulatory specialists to decide.
Use the self-assessment to prepare better questions
A self-assessment earns its place when it makes the next review sharper. Five habits help:
- Treat each question's evidence line as a request list. Every question in the assessment names the evidence to check, for example "A decision record and the exact approved evidence it references." Ask for exactly that.
- Turn each unknown into a named task. Write down who will find out, from which record, by when.
- Challenge the top answers hardest. The highest practice descriptions describe reviewing, checking or maintaining a practice over time. Ask to see the record of the last time that happened.
- Hand the list to whoever owns reviews and audits. They decide whether an internal evidence review is enough or whether something formal is due. The self-assessment does not make that call.
- Keep the profile as context, not as evidence. It explains why you looked. The records you examined are the evidence.
If you have a profile in front of you, the section of the assessment page that explains what the score, bands and prompts mean is the place to start. If you do not have one yet, the assessment runs in this order: choose a scope, answer the seven practice areas, review your answers, enter your first name, last name, work email and company, then see the profile and download it as a PDF.
What this comparison does not establish
- It does not define an audit standard or describe the audit procedure of any organization. Yours governs.
- It does not say which audits, inspections or evaluations your organization is subject to.
- The decision table describes typical characteristics. A well-run internal evidence review can be more searching than a narrow audit.
- The example is synthetic. Its numbers come from the scoring model, not from observed teams.
- The assessment's own limitations section applies to everything said here about the tool.
Related reading
- Interpret a CSV/CSA maturity profile: a worked example: the arithmetic behind area scores, coverage and possible ranges, with every answer shown.
- Why unknown answers and high scores need careful interpretation: why a score is not a benchmark and unknown is not zero.
- How to assess CSV/CSA maturity: how to collect a small evidence sample before you answer, so fewer answers rest on memory.
This article describes assessment model csa-maturity-2.0.0.
Sources and references
- S04EudraLex - Volume 4 - Good Manufacturing Practice (GMP) guidelines. European Commission.
Official index of EU GMP guidelines; listed Annex 11 as 'Computerised Systems (revision January 2011)' when checked on November 12, 2026.
- S36EudraLex Volume 4, Annex 11: Computerised Systems. European Commission.
EU GMP guideline, revision 1, with a stated deadline for coming into operation of June 30, 2011; the version linked from the EudraLex index on November 12, 2026.
