Regulations and compliance · Practical guide

Audit trail vs system log: which GxP evidence do you need?

A result change may require the original and new values, time, identity, reason and link to the record. A network log can show that a connection existed while omitting much of the information needed to interpret the change.

Table of contents

At a glance

How does an audit trail differ from a technical system log?

A technical system log may record errors, connections or operating events. A GxP audit trail should make relevant changes to data traceable in context. Assess the information actually captured rather than the name the supplier uses. A login event alone does not explain a change to a test result.

Scope and Applicability

There are different types of protocols and hybrid forms. This article outlines their documentation function. It does not claim that every event must be recorded in a single central audit trail.

Sources: [1] [2]

Different questions require different records

A system log typically records technical events such as start-up, connection, or errors. An audit trail supports traceability of relevant actions regarding data or configuration. However, the labels themselves do not guarantee content accuracy. Verify whether a specific entry addresses a concrete question: Who modified a result value, or why did the connection to the recipient fail?

A technically detailed log alone, without business context, may be insufficient for a technical investigation. An audit event, in turn, may not provide a complete diagnosis of a server failure. Both sources can be used together. What is essential is the required statement and actual coverage, not the name of the file or function.

Capture data sources, fields, time reference, access and storage. Name which changes are not logged or are only visible by another way. A general manufacturer statement "Logging exists" is not enough for your own evaluation. The proof should show concrete events.

Sources: [1] [2] [3]

Create an event and statement matrix

Assign the required information to each relevant event. For a result change, original and new values, person, timestamp, and reason may be relevant. For a transmission failure, the message, recipient, and processing status are required. These informations may be distributed across multiple sources.

Check how sources are consolidated. A common business key or another traceable reference helps. Similar time stamps alone may be insufficient given a high volume of events. The linkage should also be understandable to a later reviewer, not just to the administrator who experienced the incident.

Assign possible event causes to technical or procedural sources
QuestionPossible sourceTo be verified
Who modified a result value? Appropriate Audit Trail Person, value, and record linkage
Why did the message fail to arrive? Interface or system log Connection to the affected business object
Were the access rights extended? Evidence of the permission change Actual Usage and Professional Effectiveness
What was the active status? Configuration and Change Record Reference at the time of the event
Is the history complete? Scope and operational testing Missing paths, filters, and failures

Sources: [1] [2] [3]

Example: Login is known, result change cannot be assigned

Example shows a system log indicating that user U-3 was logged in. A result value was modified during this period. However, there is no adequate traceability to identify which specific user performed this particular change. Login alone does not establish accountability, particularly when multiple access points or technical processes are involved.

Generate a known result change in a controlled environment. Review system logs and audit data separately. Which source shows login events, and which shows the specific record and its modifications? Document the statement limits. A plausible temporal association shall not be considered definitive assignment.

Add a technical logging pathway. When a service account modifies data, the traceability of the associated processing activity must be evaluated. Not every automated change constitutes a personal signature. The architecture should specify which actions are automated and where a human decision may have preceded the automation.

Test cases for linking application submission and result modification
Test caseOwn actionExpected finding
Sign-in U-3 register Technical access traceable
Data modification Change the specified value Clear record and relevant changes identifiable
Second User Register in parallel and edit other processes No confusion solely due to similar times
Technical process Automatically trigger change Process and data traceability
Logging outage Controlled interruption logging required Missing coverage is detected and evaluated

Sources: [1] [2] [3]

Time, Integrity, and Access as Common Foundations

Timestamps require a clear reference point. Different time zones or unsynchronized systems can complicate the event sequence. Verify the required temporal alignment for your process. The requirement does not mandate universal millisecond precision. The need is driven by the required distinguishability.

Evaluate who can modify or delete logs and audit data. A technically sound record is compromised if it can be uncontrolledly removed. Technical and organizational controls can act in concert. Describe their purpose and limitations rather than making general claims of immutability.

Control access for review and investigation. A protocol may exist, but it may be practically unattainable for the responsible person. An unfiltered admin export may contain information that is not required by every reviewer. Appropriate access balances traceability and adequate information protection.

Sources: [1] [2] [3]

Organize review and technical monitoring appropriately

A technical monitoring identifies, for example, outages or recurring errors. A specialist review evaluates the significance of critical data modifications. Name both tasks and their handover point. Restoring a connection may resolve a technical disturbance, while missing results still require further investigation.

Use targeted criteria. Technical warnings are not automatically considered GxP deviations; professionally significant changes are not automatically visible in infrastructure monitoring. A thoughtful selection reduces unnecessary alerts without obscuring relevant risks. Review them against known events.

Obtain the linkage between finding, data record, and corrective action. If an error recurs, a root cause may be systemic. A single closed ticket does not address this question. The connection between IT incidents, technical assessment, and potentially CAPA should be clearly defined in the operating model.

Sources: [1] [2] [3]

Consider storage and system changes

Technical logs often have different retention periods compared to business records. Assess which information is required to support the necessary traceability. A short standard retention period should not go unnoticed in removing the sole relevant change information. Conversely, every diagnostic log does not require blanket permanent retention.

During migration or decommissioning, the context of protocols must be retained. A file lacking user assignment, record key, or interpretable timestamp may lose its evidential value. Identify specific historical investigation cases. An export should not only be readable, but remain understandable in relation to the intended query.

Sources: [1] [2] [3]

Start with the question that the evidence is intended to address

A technical log often does not capture substantive changes. Conversely, an application audit trail may fail to capture important technical events, such as a direct intervention at the database level. Therefore, evaluation must consider relevant change pathways and existing controls together. A collection of multiple logs does not constitute a complete evidence framework.

Assign a task to each review. The IT reviewer may investigate potential operational disruptions, while the technical reviewer evaluates data modifications prior to a decision. Both require appropriate information and access. Without this separation, records may be superficially reviewed twice or critical events may be entirely overlooked.

Sources: [1] [2] [3]

Draft: Investigate a change and a technical failure together

The table summarizes the professional outcome change and the technical connectivity failure. The mere co-occurrence of timing does not establish causality. Document any credible, common references and remaining uncertainties; technical recovery and the evaluation of previously processed data are separate outcomes.

Sample filling example – internal work aid
Area to investigateExample entryPurpose of the evidence
Business case Result E-90 for the specified sample. Technical basis.
Audit data Value change with required context. What was actually changed?
Technology Connection errors during the appropriate time period. What operating condition existed?
Time Sources with explicit temporal reference. Chronology without unresolved shifts.
Assignment Message and record reference. Compound rather than mere speculation.
Boundary Absence of evidence is explicitly stated. No undue conclusion.

Sources: [1] [2] [3]

Next steps

  • Define relevant data and operating events.
  • Verify the actual content of the existing protocols.
  • Assess gaps and alternative change routes.
  • Match review tasks, accesses and storage.
Further reading:

Project experience and collaboration

Correlate event records to establish what happened

The process owners and IT describe the same transaction from their respective perspectives. What changed in the business record, and which technical events were logged? We link the records through reliable identifiers and check them against known test events. A login alone does not establish that someone made a particular change. Direct technical processing may require a different source from a change made through the user interface. Your team can see which history answers which question and where a link is still missing.

Frequently asked questions about Audit trails and system logs

Is a log with a timestamp automatically an audit trail?

No. Timestamps are only part of the required evidence. Content, assignment, and protection must align with the traceability function.

Sources: [1] [2]

Can IT assume responsibility for the audit trail review?

Only to the extent that competence, independence, and process knowledge are appropriate for the specific task. Technical log review and technical data review are not equivalent.

Sources: [1] [2]

Must all logs be retained for the same duration?

Not automatically. Storage depends on their respective detection function and the applicable requirements.

Sources: [1] [2]

Can a system log assume the function of an audit trail?

This depends on the actual content and specific requirements. The name is not decisive. Review concrete data modifications, assignments, time stamps, integrity, and availability. A log containing only login entries and server errors typically does not indicate who modified which specific technical record and for what purpose.

Must technical logs and process records be kept in the same system?

Not mandatory. Multiple sources can effectively work together if relationships and access are traceable. Assess the connection based on specific events. A shared repository is not a substitute for appropriate content; separate repositories are not inherently erroneous, provided the required investigation remains practically feasible.

Why is a login not automatically considered evidence of a data modification?

It shows access, but does not necessarily indicate specific actions. Additional users or technical processes may be active simultaneously. Verify record linkage and actual change information. A plausible temporal proximity must not be assumed as definitive personal attribution without appropriate supporting evidence.

How do we handle different time zones in protocols?

Document the original time reference and, if applicable, any used conversion. Verify that the sequence required for your investigation remains understandable. Universal accuracy is not required. The requirement follows the specific event sequence and its meaning in the process.

Sources and review date

The classification is based on the listed regulatory and authority sources. Incident matrix, assignment example, and review cases are standalone support tools. Which specific events a particular product actually captures must be verified at the actual site of use.

  1. European Commission: EU GMP Annex 11, Computerised Systems (2011)Sections 9, 12, and 13: Traceability, access control, and incident assessment; log architecture is system-specific. Current version as of EudraLex on 14.10.2026; no universal project or deadline mandate.
  2. PIC/S: PI 041-1, Good Practices for Data Management and Integrity (2021)Inspection Guidance on Data Integrity; in particular electronic data, transmission, access, and review.
  3. FDA: Data Integrity and Compliance With Drug CGMP – Questions and AnswersFinal Guidance, December 2018; no blanket extension to all GxP regulatory areas.

Sources checked: .

Daniel Herrmann, Daniel Herrmann Consulting

Daniel Herrmann Consulting

Integrate events and technical statements.

We support you in assessing the current situation, clarifying responsibilities, and implementing appropriate evidence within operations. The type of support that is most effective depends on your system, your team, and the specific objective.

  • Computer system validation →

    Bring your GxP systems into validated operation, from laboratory software, MES and eQMS to ERP. We define the scope, plan risk-based tests using GAMP 5 and hand the evidence to QA for release.

Send a project enquiry Customer testimonials →