// data integrity
Data does not become reliable at the end
It becomes reliable — or fails to — the moment a system is chosen, an access profile is configured, a decision is made about where a file ends up. ALCOA+ is not a review performed afterwards: it is the criterion behind those decisions.
Nine attributes, translated into technical decisions
For each attribute: what we look at when we open up a system, and what we most often find out of place.
- Attributable
-
Every record must be traceable to whoever created or changed it, with date and time. This applies to the operator on the line as much as to the system administrator.
What we find: shared logins on shop-floor terminals, service accounts used by people, administrators without separate traceability.
- Legible
-
Readable and understandable for the whole retention period, not just today. Format matters: a proprietary file is legible only as long as the software that opens it still exists.
What we find: archives tied to discontinued software versions, exports stripped of metadata, scanned printouts standing in for the data.
- Contemporaneous
-
Recorded at the time the activity takes place. This depends on something mundane and frequently overlooked: that every system agrees on the time.
What we find: clocks out of sync between instruments and servers, inconsistent time zones, users able to change the system clock.
- Original
-
Raw data must be retained together with the metadata that makes it interpretable. A certified copy is acceptable, but it has to be managed as such.
What we find: raw data on instrument local drives with no backup; only the report retained while the original trace is overwritten.
- Accurate
-
Correct and free of undetected errors. Achieved through checks at data entry and by removing manual transcription between systems.
What we find: manual transfer from instrument to spreadsheet, free-text fields where a controlled list belongs, no limit checks.
- Complete
-
All data, including repeated analyses and discarded results, with the reason recorded. The absence of any failing result is itself a signal.
What we find: untracked trial injections, reintegrations without justification, acquisitions that can be deleted without leaving a trace.
- Consistent
-
The sequence of events must be reconstructable and consistent across the different systems describing the same fact.
What we find: interfaces silently dropping records, timestamps in mismatched formats, reconciliations never performed.
- Enduring
-
It has to survive the full retention period. This concerns media, backups and the ability to read them many years later.
What we find: backups never tested by restoring, tapes with no compatible drive left, no plan for data from decommissioned systems.
- Available
-
Retrievable and reviewable within reasonable time when needed — typically during an inspection — without depending on one individual.
What we find: archives only their author can reach, restores that take days, no written procedure for extracting data.
// how we work
From finding to fix
A list of problems helps nobody if it does not say where to start. Every gap is prioritised by risk to the patient and by how likely it is to surface during an inspection.
-
STEP 01
Gap assessment
Systems reviewed against the nine attributes, verified hands-on rather than on paper alone.
→ ALCOA+ matrix per system
→ list of gaps
→ risk classification -
STEP 02
Remediation plan
Actions ordered by priority, with effort estimates and a clear split between quick fixes and work that needs a project.
→ prioritised plan
→ interim controls
→ structural changes -
STEP 03
Implementation
Configuration, procedures and training. Changes to validated systems go through change control.
→ configuration
→ operating procedures
→ user training -
STEP 04
Periodic review
Integrity erodes over time. Audit trail review has to be made sustainable, or it stops being done.
→ audit trail review
→ access review
→ restore verification
On audit trail review
This is the requirement most often written into a procedure and never carried out. The reason is almost always practical: the audit trail produces thousands of lines and nobody has time to read them.
The answer is not to demand more discipline from people, but to reduce the volume to be reviewed: filters on critical events, risk-based review, automatic alerts on the actions that actually matter — deletions, changes to approved data, configuration changes.
A procedure people can follow is worth more than a perfect one they ignore.
Where does your data stand?
An ALCOA+ gap assessment on a bounded system tells you where you actually are within a few weeks, without stopping production.