Regulatory mapping¶
Validance's run records, audit trail, and workflow identity are built to line up with specific clauses of the FDA and EU rules for computerised systems used in regulated environments. Validance is not itself subject to those rules; its customers are. This page maps each Validance deliverable to the regulatory text it corresponds to, and says where the correspondence stops.
This page is a technical reference for QA, regulatory affairs, and validation teams evaluating Validance for use in their environment. It is not legal advice.
Summary¶
| What you owe | What Validance produces |
|---|---|
| Computer-generated, time-stamped, secure audit trail (21 CFR Part 11 §11.10(e)) | Cryptographically linked, fixed-format-timestamped record per execution, with previously recorded values preserved |
| A system-generated record of GMP-relevant changes and deletions, available and convertible to an intelligible form (EudraLex Annex 11 §9) | The same record, retrievable as JSON by API, with per-entity verification so any specific workflow or task history is independently checkable |
| Recorded identity of the operator entering, changing, confirming or deleting data (Annex 11 §12.4) | Not provided today. Events carry a system actor, not a named person. See Operator identity |
| Digital records as objective evidence (FDA CSA 2026 §V.A.6) | System logs, audit trails, file content hashes generated by the software, structured for direct use in assurance records |
| Versioned, hash-identified workflow definition under controlled change (Annex 11 §10) | Content-addressable workflow with SHA-256 definition hash; change is detectable by hash diff |
| Per-run forensic record on demand, without manual reconstruction | Produced as a side effect of execution, retrievable by API at any time |
| ALCOA+ as a structural property of records | See ALCOA+ mapping below |
Detail¶
Audit trail (21 CFR Part 11 §11.10(e))¶
Part 11 §11.10(e) requires "use of secure, computer-generated, time-stamped audit trails to independently record the date and time of operator entries and actions that create, modify, or delete electronic records. Record changes shall not obscure previously recorded information."
Validance generates audit records automatically for every workflow event. Each record contains an actor, a timestamp, an event type, and event-specific details. Records are linked cryptographically; modification is detectable. Audit emission is part of the execution path, so a successful run necessarily has a recorded audit trail. See Audit and evidence.
The clause says "operator entries". Validance records the entry and its time; it does not yet record a named operator. See Operator identity.
Audit trail (EudraLex Annex 11 §9)¶
Clause 9 reads, in full:
"Consideration should be given, based on a risk assessment, to building into the system the creation of a record of all GMP-relevant changes and deletions (a system generated 'audit trail'). For change or deletion of GMP-relevant data the reason should be documented. Audit trails need to be available and convertible to a generally intelligible form and regularly reviewed."
Three sentences, three different owners.
The third is Validance's. Audit records are persistent, retrievable as JSON by API, and independently verifiable, so they are available and convertible to a generally intelligible form, and they are structured for QA review and inspection response.
The second is yours, with one exception. For a change or deletion of data, and for a change to a workflow definition, Validance records that it happened and when, but has no field for why, so the documented reason has to come from your own procedure.
The exception is an approval decision, which does carry a reason: resolving a gate accepts a free-text reason and stores it with the decision and its timestamp, and it is retrievable with the approval. Read it as a recorded rationale, not as attribution. The accompanying decider field is a string the caller supplies, and nothing authenticates it, so it records who the caller said decided rather than who did. See Operator identity.
The first is yours as well: whether a given workflow's data is GMP-relevant, and what the risk assessment concludes, is an intended-use determination Validance cannot make for you.
Operator identity¶
The sentence QA readers most often expect to find under clause 9 is in clause 12.4 instead, under Security:
"Management systems for data and for documents should be designed to record the identity of operators entering, changing, confirming or deleting data including date and time."
Validance does not satisfy this today, and the gap is worth stating precisely rather than softening.
Audit events carry an actor field, but its value is a system identity rather than a person: the engine sets it according to the entry path, so API-triggered work is attributed to the API, not to whoever called it. Triggering a run accepts no caller identity at all.
One surface is different, and it needs the sharper warning rather than the softer one. Resolving an approval accepts a decider, which is stored with the decision and appears in the audit trail. But it is a free string, nothing authenticates it, and it falls back to the literal api when omitted. The decider name is unconnected to the credential that made the call, so any caller who can reach the endpoint can resolve a gate and have the trail name whoever they typed. Treat that field as a claim recorded, never as attribution established: it is weaker than an empty field, because it reads like evidence.
In practice: the when, the what, and the on which data of clause 12.4 are recorded. The who is recorded at system granularity only. If your intended use needs named-operator attribution inside the engine's own records, Validance is not sufficient for that on its own today. Either the system that fronts Validance has to hold that attribution, or it is carried as an open item in your validation. Verified operator identity is on the roadmap.
Workflow definition under change control (Annex 11 §10)¶
Annex 11 §10 requires that "any changes to a computerised system including system configurations should only be made in a controlled manner in accordance with a defined procedure".
Validance's content-hash workflow identity makes change detection mechanical: comparing two definitions reduces to comparing two SHA-256 hashes. Combined with version control of the JSON files and the registration step that issues a hash, this provides a substrate for a documented change control procedure.
Digital records over paper (FDA CSA 2026 §V.A.6)¶
The FDA Computer Software Assurance guidance (3 February 2026) recommends "incorporating the use of digital records, such as system logs, audit trails, and other data generated and maintained by the software, as opposed to paper documentation, screenshots, or duplicating results already digitally retained by the software when establishing the record associated with the assurance activities."
Validance's run records, audit trail, lineage queries, and verification API are designed to be used directly as objective evidence in assurance documentation, without manual transcription.
Per-run forensic record¶
For an orchestration engine in a regulated environment, the per-run forensic record is a primary deliverable: workflow definition hash, engine version, container image hashes, file content hashes, lineage, actor, and timestamps for every execution, retrievable on demand. Validance produces this as a side effect of execution; no manual reconstruction is required.
ALCOA+ mapping¶
ALCOA and ALCOA+ are the data-integrity principles inspectors apply to electronic records. The FDA's data-integrity guidance uses this vocabulary, and the mapping below is against that guidance.
| Principle | How Validance addresses it |
|---|---|
| Attributable | Every audit event records an actor, a timestamp, and the entity it refers to. The actor is a system identity, not a named operator: see Operator identity |
| Legible | Audit records are JSON, human-readable on retrieval, and remain so through the engine's lifetime |
| Contemporaneous | Audit emission is on the execution path; events are recorded at the time they occur, not reconstructed |
| Original (or true copy) | File content hashes preserve the original artefact; the run record references the originals by hash |
| Accurate | Every input and output file's content hash is recorded; any change in content produces a different hash |
| + Complete | Failed runs and aborted runs are captured with their failure mode; records include all events, not just successful ones |
| + Consistent | Audit records carry fixed-format timestamps and chronological ordering |
| + Enduring | Records are persistent in the engine's backing database; archival and retention beyond engine lifetime are operational concerns |
| + Available | All evidence is retrievable by REST API, by run, by workflow, or globally |
What remains your responsibility¶
Validance produces the runtime substrate for compliance evidence. The following remain the customer's responsibility:
- Per-workflow risk assessment and qualification
- Authoring of validation documents (validation plans, qualification protocols, summary reports)
- Electronic signatures, which stay in your quality system. Validance has no e-signature feature and no e-signature integration
- Documented reasons for change or deletion of GMP-relevant data, and for changes to a workflow definition (Annex 11 §9). Approval decisions are the exception: those carry a reason
- Named-operator attribution, where your intended use requires it (Annex 11 §12.4), including any control that would make a recorded approver verifiable rather than asserted
- Periodic review, supplier qualification, training, business continuity
- Classification of the workflow's intended use under the relevant predicate rule
These activities are bounded and procedural. Validance is the substrate they rest on; it does not replace QA.
Scope of this page¶
The clauses mapped above are the pharma-side GMP and FDA records rules. If you are in in-vitro diagnostics or medical devices, the frameworks that govern you are not mapped here yet. The engine properties are the same; the clause-level mapping is work we have not published. Ask and we will go through it directly.
Source references¶
| Reference | Where to read it |
|---|---|
| 21 CFR Part 11 | eCFR |
| FDA Computer Software Assurance for Production and Quality Management System Software (February 2026) | FDA |
| FDA Data Integrity and Compliance With Drug CGMP — Q&A (2018) | FDA |
| EudraLex Volume 4, Annex 11 — Computerised Systems | EU Health |