Review Validation Traceability with One Evidence Chain
Review validation traceability by following one need to its rationale, evidence version, findings and disposition, then again after a change. Includes a synthetic chain with a deliberately broken link.
Written by CSV/CSA Maturity Editorial TeamPublished 11 min read
On this page7 sections
A validation traceability review does not need the whole matrix. Choose one need that matters and follow it by hand: to the reasoning that set the assurance effort, to the exact evidence version that was relied on, to any finding and its disposition, to the release decision, and then through the most recent change. If each step can be taken from controlled records without asking the author, the trace is meaningful. If a cell is filled in but its target cannot be retrieved, or is not the version that was approved, the link is populated and the trace is broken.
This how-to is for validation and assurance leads working in computer system validation (CSV) or computer software assurance (CSA). CSV here means validation, not comma-separated values. The sequence works on any record medium, whether the links live in a spreadsheet matrix, a document management system, a lifecycle tool or paper binders with an index. It reviews one chain as a sample. It is not an audit, and it does not show that every chain is sound.
A populated link is not yet a meaningful trace
A populated link is a reference in a cell: a need ID points at a test ID. It shows that someone made a connection at some time. A meaningful trace is a connection a reviewer can use today. Five checks separate the two.
| Check | Populated link | Meaningful trace |
|---|---|---|
| Can the target be retrieved? | A reference exists. | An authorized reviewer can open the referenced record from its controlled location. |
| Is the version known? | An identifier with no version, or the word "latest". | The identifier and version match the one approved for the decision. |
| Does the evidence address the need? | The test title resembles the need. | The executed steps or reviewed evidence exercise the stated need and its acceptance expectation. |
| Are findings connected? | Issues sit in a separate log. | Findings and their dispositions can be reached from the chain. |
| Has the link survived change? | Set once at release. | Reassessed when the need, the evidence or the system changed. |
A matrix is one way to hold these links. Others include references inside a lifecycle tool, a cross-reference section in a validation summary and annotated requirement documents. The format matters less than whether the relationship is usable and controlled. Generating a matrix shows that links were written down. It does not show that the evidence behind them is adequate.
What the regulatory sources say about traceability
Two sources give context, each for its own setting. For medicinal products in the EU, Annex 11 says user requirements should be traceable throughout the life cycle (clause 4.4) (Annex 11 [S36]). As of November 12, 2026, the EudraLex Volume 4 index [S04] lists Annex 11 as the January 2011 revision.
FDA's final guidance on computer software assurance, issued on February 3, 2026, gives nonbinding recommendations for software used as part of medical device production or the quality management system. It does not recommend a traceability matrix and mentions traceability only in passing (FDA guidance, PDF [S03]).
Our reading is that neither text fixes a format for showing traceability. What you must keep is decided by your own procedures and the rules that apply to your activities.
What "a retrievable current evidence version" means
Question Q09 of the assessment asks how reliably the team can identify the evidence version used for an assurance decision. Its guidance says to evaluate control and retrievability, not whether the evidence lives in a particular platform. Put together, the question is whether you hold a retrievable current evidence version. That phrase has four parts.
- Evidence is the record that supports the decision: an executed test record, a review record, a supplier document assessed against your intended use, or a system log extract.
- Version is an identifier that separates this state of the record from earlier and later ones. A document number with a revision, an execution run ID, or a signed and dated paper original with a stated storage location all qualify. A file name ending in "final" does not, because nothing stops two files from carrying it.
- Current means the version that supports the system as it is released now. After a change, the superseded version is kept as history and the trace points to the one that replaced it.
- Retrievable means a reviewer with the right access can obtain that exact version from its controlled location, without a personal copy and without the one person who remembers where it is.
Three questions test this quickly. Which version did the approver see? Is that the version the trace points to? Can it be opened now, from the controlled location, with its approval status visible? If any answer is "probably", choose a lower statement for Q09, or "I do not know yet" until someone checks.
| Q09 answer | Statement in the model | What a reviewer finds |
|---|---|---|
| 0 | "The exact evidence version is often uncertain." | References without versions, and no reliable way to tell which copy was approved. |
| 1 | "Naming conventions help, but conflicting or uncontrolled copies remain." | File names suggest a version, but more than one candidate exists outside control. |
| 2 | "The approved evidence version and its relationship to the decision are identifiable." | The decision cites an identifier and version, and that version opens from the controlled location. |
| 3 | "Checks detect broken links, superseded copies or incomplete evidence packages before reliance." | A routine check finds version gaps before a decision depends on them. |
The review sequence in six steps
1. Choose one important need
Pick a need that would matter to product quality, patient safety or the integrity of a regulated record if it were not met. Where possible, pick one that has been through a change since release. Do not pick the need you already know is tidy.
Write down its identifier, its wording and its acceptance expectation as they stand in the approved record. If the wording is too vague to say what evidence would satisfy it, stop and fix that first. Writing needs with observable acceptance expectations covers how.
2. Find the supporting rationale
Locate the reasoning that decided how much assurance this need received: the risk assessment entry, the assurance rationale or the relevant part of the test strategy. Check that it refers to this need or its function, not only to the system in general, and note its identifier and version. If no rationale can be found, record that as a gap in the chain. Do not infer one from the tests that happen to exist.
3. Locate the evidence and its version
Follow the link from the need to the evidence. Record the identifier and version the trace gives you, then open the record from its controlled location and compare. Check the approval status and date of that version.
Then look for other copies: an earlier revision, a later one, an export on a shared drive. More than one candidate is not automatically a problem. Not knowing which one the decision relied on is.
4. Inspect issues and their disposition
Read the evidence for findings: failed steps, deviations, observations and open questions. For each one, find the issue record, its owner, the impact assessment and the disposition, and check that the disposition also refers to evidence by version. A chain with a justified, approved disposition is intact. A chain whose finding leads nowhere is not, even if every test eventually passed.
5. Confirm the release context
Find the decision that put the system, or this function, into use: a release approval, a validation summary conclusion or a change closure. Check which evidence versions it cites and whether its date falls after their approval. A decision dated before the evidence it relies on was approved needs an explanation. So does a decision that cites evidence without a version.
6. Repeat after a change
Take the most recent change that touched the need, its function or its evidence. Check that the impact review named this need, that affected evidence was re-executed or re-reviewed where the review said it should be, and that the trace now points to the resulting version. This step turns a one-time check into a test of whether links are maintained. For the change record itself, see how to review a software change or supplier release.
Worked chain with a deliberately broken link
The sample comes from an invented calibration-scheduling system. The reviewer follows two needs after change CHG-031, which altered role permissions.
| Chain element | Need UR-014 | Need UR-021 |
|---|---|---|
| Need ID and statement | UR-014: an instrument whose calibration is overdue cannot be selected for use. | UR-021: changing a calibration interval requires a reason and approval by an authorized role. |
| Assurance decision | RA-006 v2: a failure could let an overdue instrument be used. Scripted tests at the due-date boundary, with independent review. | RA-006 v2: a failure could let an interval be extended without oversight. Scripted negative test for each role, plus review of the audit trail entry. |
| Evidence version | TE-0085 v1, approved, in the controlled repository. Release decision REL-005 cites it with its version. | The trace cites "TE-0088" with no version. The repository holds v1, approved before release, and v2, approved when CHG-031 closed. REL-005 also cites "TE-0088" without a version, and a file named "TE-0088 final" sits on a shared drive. |
| Finding | None recorded. | ISS-012, raised after release: a supervisor role could change an interval without a second approval. |
| Disposition | Not applicable. | ISS-012 closed as "corrected under CHG-031 and retested". The retest reference reads "TE-0088", with no version. |
| Change impact | The CHG-031 impact review lists UR-014 as not affected and says why. No evidence change needed. | The CHG-031 impact review lists UR-021 as affected and calls for a retest. The trace was not updated when the change closed. |
Every cell in the UR-021 column is filled in, and a count of populated links would report the need as traced. The reviewer still cannot establish three things from the trace: which version of TE-0088 each reference means, whether the closure of ISS-012 rests on the retest in v2, and what the copy on the shared drive is.
Record the uncertainty before resolving it
The reviewer writes down what the records show and what they do not. Two shortcuts are avoided: assuming the latest version is the right one, and editing the trace so that it looks complete.
| Field | Entry (synthetic) |
|---|---|
| Chain reviewed | UR-021, after CHG-031 |
| What the records show | Two approved versions of TE-0088 exist. The trace, REL-005 and ISS-012 cite the test without a version. |
| What cannot be established yet | Which version each reference means. Whether the shared-drive file is a controlled record. |
| What is not being concluded | That the function is wrong, or that the retest was not performed. |
| Owner and action | Validation lead (role) confirms versions from the approval history. Record owner (role) assesses the shared-drive file. |
| Interim position | UR-021 is reported as "link not verified", not as traced. |
Resolve it from controlled records
The owner goes to the approval history, not to memory. In this sample it shows that v1 was approved before REL-005 was signed, so the original release relied on v1. It also shows that v2 added the supervisor role, was executed after the permission change and was approved when CHG-031 closed. The shared-drive file is a working export of v1 with no approval record.
The corrections follow from those facts:
- The trace for UR-021 is updated to cite TE-0088 v2 as the current evidence and v1 as superseded history.
- The closure of ISS-012 is amended, under the record-correction procedure, to cite TE-0088 v2.
- REL-005 is not rewritten. Under the same procedure, a dated note links it to v1, the version it relied on.
- The shared-drive file is removed or marked as an uncontrolled copy, as the record owner decides.
- Because the cause was a change closing without a trace update, the reviewer samples two other needs touched by CHG-031 and proposes a closure step that requires versioned evidence references.
Had the approval history not shown which version was relied on, the entry would stay open as "cannot be established", and the organization's quality process would decide what is needed, for example re-execution or a documented justification. Filling the cell with the most likely answer is not a resolution.
How the chain review maps to the assessment
In model version csa-maturity-2.0.0, the traceability practice area has three questions, each answered with one of four practice statements scored 0 to 3 or with "I do not know yet".
| Question | What a 2 describes | What the chain review tells you |
|---|---|---|
| Q08 | "Important needs and risks are connected to the relevant assurance activities and evidence." | Steps 1 to 3: whether the need reaches a rationale and evidence that address it. |
| Q09 | "The approved evidence version and its relationship to the decision are identifiable." | Steps 3 and 5: whether the version relied on can be named and opened. |
| Q10 | "The change process identifies and updates the evidence relationships that are affected." | Step 6: whether links were updated when the change closed. |
As found, the synthetic sample reads more like the statements scored 1 for Q09 and Q10: uncontrolled copies remain, and links are updated when a reviewer asks. One chain is too small a sample to answer for a whole scope, but it shows which statement to test next.
If you cannot yet say which version a decision relied on, "I do not know yet" is the accurate answer, and the model does not count it as zero. In a synthetic profile where every answer is 2 except Q09 marked unknown, numeric coverage is 94.4%, the traceability area shows a possible range of 44.4 to 77.8 in place of a score, and the overall result is a possible range of 63.4 to 68.3 with no band. The scoring methodology shows the arithmetic.
You can answer the traceability questions for your own scope. The assessment asks for a scope, takes you through the seven practice areas and a review step, then asks for your first name, last name, work email and company before it shows the profile, which you can download as a PDF. The result is a self-reported profile for planning, not an audit or a certification.
What this review does not establish
- One chain is a sample. A sound chain does not show that the other needs are traced, and a broken one does not show that they are not.
- A complete trace shows that evidence can be found and matched to a decision. It does not show that the testing was sufficient, that the risk reasoning was right or that the function works.
- The review makes no regulatory finding. Whether a gap matters under the rules that apply to you is a question for your quality and regulatory specialists.
- The chain layout is one option among several. Nothing here is a required template.
Related reading
- Document intended use and clear acceptance expectations: a vague need gives a trace nothing definite to point at, so start here if step 1 stalls.
- Software changes and supplier releases: the evidence to keep: the change record that step 6 depends on.
- Connect functional risk to an assurance approach: the rationale that step 2 looks for, with a worked decision record.
This article describes assessment model csa-maturity-2.0.0.
Sources and references
- S03Computer Software Assurance for Production and Quality Management System Software: Guidance for Industry and Food and Drug Administration Staff. U.S. Food and Drug Administration.
Final guidance document (PDF), issued February 3, 2026; nonbinding; supersedes the final guidance issued September 24, 2025.
- 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.