What Data Integrity Controls Should Buyers Audit at a Dispensing Supplier?

Data integrity is different from data retention

Retention asks whether a record still exists. Integrity asks whether the record is trustworthy. A supplier can keep five years of CSV files that lack units, user identity, revision history or product linkage. That archive may be large and nearly useless during a complaint.

Start with the decisions data supports: release, containment, deviation, CAPA, customer notification and process validation. The stronger the product risk and the harder the defect is to detect, the stronger the control expected around its evidence.

Audit user identity and access rights

List roles that can run recipes, edit limits, acknowledge alarms, release product, modify inspection results, administer accounts or delete files. Compare approved rights with live accounts. Look for shared credentials, generic administrator users, departed employees and unnecessary privileges.

Unique user identity matters because an audit trail saying 鈥淎dmin changed ratio limit鈥?does not show who made the decision. Privileged access should be limited, reviewed and monitored. Emergency access needs its own controlled process and post-use review.

• Sample active accounts against current employees and roles.
• Review failed logins and privilege changes.
• Check password/session rules appropriate to the environment.
• Confirm production operators cannot erase audit evidence.
• Verify account removal after role change or departure.

Challenge recipe and parameter change controls

Select a critical recipe and compare the approved master, machine version and audit history. Make sure the revision active during sampled production can be reconstructed. Review changes near complaints, alarms, maintenance or yield shifts.

Where the controller lacks a suitable audit trail, compensating controls may include restricted access, controlled download, checksum comparison and witnessed revision logs. The limitation should remain visible in the supplier risk assessment.

Inspect alarm, override and deletion history

Review whether alarms remain in history after acknowledgement and whether users can delete or filter them from exported reports. Critical alarms should link to product hold, investigation and restart. Overrides, disabled interlocks, manual mode and bypassed inspections need equal visibility.

Ask for an unfiltered export covering a normal period and compare it with the management summary. Gaps in sequence numbers, abrupt file resets or recurring alarms absent from reports deserve investigation. A clean dashboard can result from filtering rather than a stable process.

Verify timestamps and system clocks

Machine, PLC, robot, vision, MES, quality and warehouse systems may use different clocks or time zones. Even a few minutes of drift can make it difficult to map an alarm to affected units on a fast line. Check time source, synchronization frequency, daylight-saving handling and permissions to alter time.

Run a cross-system event comparison: choose a known startup, alarm or material scan and compare timestamps. Document any controlled offset. A timestamp without time zone or synchronized context is weaker evidence than it appears.

Control manual entries, corrections and retests

Manual data is not automatically unreliable. It needs attribution, reason, date and preservation of the original value. Paper corrections should remain legible; electronic corrections should create an audit trail. Avoid spreadsheets where formulas, limits and old entries can be overwritten without history.

Application scenario matrix

Test backup, restoration and archive readability

Review backup scope, frequency, monitoring, separation and access. Then request evidence of an actual restoration test. A successful backup job does not prove files can be restored or interpreted. Include machine controllers and local inspection computers that may sit outside corporate IT backup.

Retrieve an older archived lot, including recipe, alarms and inspection. Confirm units, field definitions and software compatibility. Before controller or software replacement, export and validate critical records in a durable documented format.

Test export completeness and report logic

Compare the on-screen record, database/source export and customer-facing report. Confirm filters, calculations, time windows, rounding and excluded statuses. Reports should identify their query logic or controlled revision. Otherwise two users may produce different 鈥渙fficial鈥?results.

For high-risk data, preserve raw files and a human-readable rendering. Hashes or checksums can help detect later file changes, but they do not prove the original input was correct. Combine technical integrity controls with process and approval controls.

Run an end-to-end evidence challenge

Select one finished serial and trace backward through shipment, release, test, recipe, alarms, material and operator. Then select one changed recipe or critical alarm and trace every affected product forward. Compare electronic evidence with physical stock and labels.

This challenge should integrate with the dispensing supplier traceability drill. The companion data-retention guide defines which record classes should remain available.

Classify findings by product risk

Do not average away an integrity failure using a supplier score. One severe evidence issue can invalidate otherwise attractive KPI results.

Cross-check interviews against live behavior

Interview operators, technicians, quality reviewers and system administrators separately, then compare their explanations with the configured system and sampled records. Ask how they correct an entry, respond to an alarm, restore a backup and handle a temporary administrator account. Demonstrations often expose informal steps that procedures omit.

Observe one real transaction from data creation through review and release. Record any paper note, local file, USB export or manual transfer used between systems. These unofficial bridges deserve control because they can break attribution, timing or completeness even when each formal system appears compliant.

Put controls into supplier governance

Define record ownership, access rights, audit trails, clock control, retention, restoration, export and notification before system migration in the supplier quality agreement. After a complaint, verify the controls as part of the CAPA effectiveness audit.

ISO 9001:2015 provides a quality-management framework relevant to documented information, traceability, monitoring and corrective action. Industry, customer, contractual, cybersecurity, privacy and regulatory requirements may be more specific.

Frequently asked questions

Does an electronic system automatically ensure data integrity?

No. Access, audit trails, configuration, time, backup and operating practices determine trustworthiness.

Are shared machine logins always unacceptable?

They weaken attribution. Where technically unavoidable, use controlled compensating records and a migration plan.

Should buyers request raw data?

For critical claims and investigations, adequate source evidence is important, subject to contract and confidentiality.

Can a screenshot prove a machine setting?

It can support evidence but usually lacks revision, history and product linkage on its own.

What if old data requires obsolete software?

Maintain validated access or migrate to a readable format before retiring the software.

What should buyers send OBO Precision?

Send the process, material, record map, controller/export limits, complaint risk and identified data gaps.