A traceability matrix that has never shown a gap is not reassuring. It is a document nobody has stress-tested. A matrix maintained by hand reports whatever it was last told, and what it was last told was usually that everything is covered.
This is the most consequential and least discussed difference between validation approaches: whether traceability is a document produced from the work, or a derived view of the work that cannot disagree with it.
How a static matrix decays
Nobody falsifies a traceability matrix. It decays through a sequence of individually reasonable steps.
A requirement is reworded during review. The test case it points at is not updated, because the rewording did not change the intent — until six months later when it turns out it did. A test case is split in two for practical reasons, and the matrix keeps the original identifier. A result is superseded by a re-execution, and the matrix cites the first one. A requirement is descoped and the row stays. A new requirement is added late and the matrix is updated at the next milestone, which is after the summary report was drafted.
Each step is small. None is wrong enough to raise in a change control meeting. After eighteen months the matrix is immaculate and increasingly fictional, and the only person who finds out is whoever is in the room when an inspector picks a requirement at random and asks to follow it through.
Derived rather than maintained
The alternative is not a better-maintained document. It is to stop maintaining a document at all.
If requirements, risk assessments, test cases and results are held as connected records rather than as separate files, the traceability is a view over them. It cannot be out of date, because there is nothing to update — rewording a requirement, splitting a test, superseding a result or descoping a requirement changes what the view shows immediately.
The important consequence is that gaps become visible while they are cheap. A coverage gap found in week three of a project is a task. The same gap found during an inspection is a finding, and it arrives with a question about what else you have not noticed. Live RTM is how we do this.
The three gaps worth watching
Inspectors look for the same three things, and they should be visible to you first.
- Coverage gaps. A requirement with no test, or no passing result. The obvious one, and the one a static matrix is most likely to hide because the row exists and points somewhere.
- Orphan tests. A test case tracing to nothing, usually left behind after a requirement was descoped. Harmless in itself, and a reliable indicator that the matrix is not derived from anything.
- Requirements covered only by not-applicable or inconclusive results. The subtlest and the most serious. The row is populated, the matrix reports coverage, and the evidence does not demonstrate anything.
The third is worth a specific check on any estate you have inherited, because it reports as green.
Why this matters more under CSA
The FDA's Computer Software Assurance guidance, finalised in September 2025 and reissued in February 2026, makes assurance depth follow per-requirement risk. That means requirements no longer map uniformly to scripted tests — one may be covered by vendor evidence, another by an unscripted test record, another by a documented rationale for no testing.
A spreadsheet with a requirement column and a test case column cannot represent that honestly. It has nowhere to put the risk determination, the assurance path chosen, or the reasoning. So organisations either keep a second document that holds the reasoning and immediately diverges, or quietly revert to scripted testing because the paperwork supports it.
Traceability that carries requirement, risk, assurance path, evidence and rationale as connected fields is what makes a risk-based approach documentable. AI Risk Assessment and CSA services are the two halves of that.
The test to run on your own estate
Pick a system nobody has prepared. Pick one requirement at random. Ask to see the risk assessment that covers it, the assurance approach chosen and why, the evidence, and whether the evidence is the current version.
If that takes more than a few minutes, the matrix is a document rather than a view. That is not a crisis and it is worth knowing before somebody else finds out. The same exercise across five requirements gives you a reliable read on an estate without a full assessment.
Validation metrics that matter covers the other measures worth tracking, and book a demo if you want this run against a live system.
Where to go next
Explore Cybroscape GxP Copilot for AI-native validation, Cybroscape TraceDraft for source-traceable clinical documentation, or book a demo to see either on your own data.