Change Management Audit Evidence: Request, Approval and Test Records

Change management audit evidence: a complete population, the request–approval–test–rollback trail, emergency change records and system log–ticket reconciliation.
Technical summary
Technology
ISO/IEC 27001:2022 A.8.32, A.8.9, ITSM, Git, Windows Update, Cisco IOS archive
Topic
Audit evidence and reconciliation for the IT change process
Level
Intermediate–advanced
Prerequisites
  • A defined change management procedure
  • A ticketing (ITSM) system or equivalent record
  • Read-only access to servers and network devices
In this article8 sections

By opening a single change ticket, the auditor can see both the design and the operation of the process; that’s why change management is sampled in almost every ITGC and ISO 27001 audit. A ticket having been opened isn’t enough. Records must show that the change was requested, its risk assessed, approved before implementation, tested, and that there was a rollback plan. Below we describe the order of preparation for the IT team, from preparing the change population to reconciling tickets with system records.

How does the auditor test the change process?

In ISO/IEC 27001:2022 Annex A, A.8.32 (Change management) requires changes to information processing facilities and systems to be subject to change management procedures; A.8.9 covers configuration management and A.8.31 separation of development, test and production environments. In ITGC audits, program changes are a separate control area; for the general framework, see the ITGC controls guide.

The auditor’s approach usually has three steps:

  1. Population request: A list of all changes in the audit period is requested.
  2. Attribute testing: For each sample selected from the list, evidence of request, approval, testing, implementation and closure is checked one by one.
  3. Reverse testing: A change that happened on the system side (patch, configuration, code deployment) is picked and its counterpart is looked for in the ticketing system.

The third step tests the completeness of the population. Unprepared teams get stuck here most: the ticket list looks clean, but a change with no ticket turns up in the switch’s configuration archive. The article control effectiveness testing explains the difference between design and operating effectiveness in detail.

Completeness of the change population

The population is all changes made to in-scope systems during the audit period: including those completed successfully, those that failed and were rolled back, and those cancelled. Providing only “closed” tickets gives the auditor an incomplete picture. If a single change not on the list turns up later, the reliability of the whole population is questioned.

The export should contain at least these fields: ticket number, change type, affected system, requester, approver and approval time, planned and actual implementation time, implementer, status and closure time. Keep the filter you used (date range, category, system) with a screenshot or the query text; the auditor must be able to reach the same number with the same filter. You can find how sample size is set relative to the population in the sample selection in internal audit guide.

Expected evidence by change type

The standard, normal and emergency distinction, also used in the change enablement practice of ITIL 4, means different evidence is expected for each type. If these types aren’t defined in the procedure, during an IT compliance audit the auditor looks for every change at the evidence level of a normal change. You end up having to document hundreds of monthly patches one by one like normal changes.

Type Definition Approval Expected evidence
Standard Low-risk, recurring, predefined The template is approved once Approved template, implementation record
Normal Planned change requiring risk assessment Approval by an authorized person or board before implementation Risk, approval, test, implementation and rollback plan, closure
Emergency Can’t wait due to a service outage or critical vulnerability Fast-track authorized approval Justification, trail of prior approval, subsequent review

The chain the auditor wants to follow for a normal change:

  • Request: What changed and why; the requester and date.
  • Risk and impact: Affected systems, need for downtime, security impact.
  • Approval: The approver’s name and timestamp; the timestamp must be before the implementation time.
  • Test: Test environment result or acceptance test record; who tested and what the result was.
  • Implementation: Actual date and time, the implementer, the version or package used.
  • Rollback: Planned rollback steps; the record if they were applied after a failure.
  • Closure: Post-implementation verification and, if needed, a review note.

Evidence of approval and segregation of duties

The essence of segregation of duties is simple: the person who develops or implements a change doesn’t approve their own change. Before the audit, filter the export and find for yourself the records where the approver and implementer fields are the same person. For software changes, protected branch and mandatory code review settings on Git-based platforms enforce approval at system level; that’s why they’re strong evidence. Present a screenshot of the setting or its configuration output together with the approval record of the pull/merge request.

The auditor also asks who has permission to make changes to the production environment. If developers have direct write access to production, the process can be bypassed. Whatever the procedure says. In that case a compensating monitoring control is needed; for example a periodic, independent review of production changes.

Reconciling system records with tickets

Reconciliation means comparing the ticket list with records showing what actually changed in the system. Prepare for the reverse test yourself: at the end of the period, extract the change trail from critical systems. The commands below can be used to get this trail on different platforms:

# Windows: successfully installed updates (WindowsUpdateClient, event 19)
Get-WinEvent -FilterHashtable @{LogName='System'; ProviderName='Microsoft-Windows-WindowsUpdateClient'; Id=19; StartTime=(Get-Date '2026-07-01')} |
  Select-Object TimeCreated, Message | Export-Csv .\WU_SRV01_2026H2.csv -NoTypeInformation

# Debian/Ubuntu: package installs and upgrades (including rotated logs)
zgrep -hE ' (install|upgrade) ' /var/log/dpkg.log*

# RHEL family: dnf transaction history (for details: dnf history info <transaction-id>)
dnf history list

# Git: merges within the period
git log main --merges --since="2026-07-01" --until="2026-12-31" --date=iso --pretty=format:"%h;%an;%ad;%s"

# Cisco IOS: configuration change archive ("archive / log config / logging enable" must be enabled beforehand)
show archive log config all

A mistake you may come across on the Cisco side is that configuration logging under archive was never enabled; then there’s no retrospective trail and reconciliation can’t be done. Match the outputs to the ticket list by date and system name, and classify every record that doesn’t match:

System record Ticket counterpart Result
SRV01 monthly security updates Standard change template OK; linked to the template
VLAN added on the core switch Normal change ticket OK; approval before implementation
Service configuration on the ERP server No ticket Deviation; explanation and corrective action required

This table is an example working template. If you found the change without a ticket before the audit and recorded its root cause and corrective action, you’re in a much better position than if the auditor finds it. For how server patches are planned, the server patch management guide is a complement. For tracking configuration baselines, NIST’s SP 800-128 is a good reference for security-focused configuration management.

Retrospective approval of emergency changes

An emergency change isn’t skipping the process, it’s speeding it up. Four elements must be defined in the procedure: which situations count as emergencies, who can give emergency approval, through which channel the approval is recorded (ticket, logged messaging, email confirmation after a phone call) and within how many working days full documentation and review are completed.

The auditor also looks at how the number of emergency changes trends across periods. If the impression is that the emergency category is used to cover a lack of planning or to avoid waiting for board approval, this can turn into a finding about the design of the process. If there are emergency changes every week, the problem isn’t an emergency. Adding the periodic list of emergency changes and their review dates to the change board minutes is the simplest answer to this.

Common findings

  • Cancelled and failed changes are missing from the population.
  • The approval timestamp is after the implementation time.
  • The approver and implementer are the same person, with no compensating control defined.
  • The test record is just the phrase “tested”; who, where and with what result isn’t clear.
  • The rollback plan is the same boilerplate text on every ticket.
  • Change logging is disabled on network devices; reconciliation can’t be done.
  • The retrospective review of emergency changes wasn’t done.

MES, PLC and recipe changes in the production environment carry different risks and must be handled separately; this is covered in the article change management in production systems. To prepare your IT change process for audit, see the change, configuration and patch management and ITGC and audit readiness service pages.

Frequently asked questions

Where does the auditor pick the change sample from?

From the change population the organization provides, that is, the list of all changes in the audit period. Showing that the population is complete is the organization’s responsibility; that’s why the auditor often tests in the reverse direction too, asking for the ticket-system counterpart of a change picked from system records.

Do we need to open a change record every time for automatic patches?

Usually not. The monthly patch cycle can be defined as a pre-approved standard change. In that case the evidence is the approval of the standard change template, the patch deployment policy (pilot and production rings) and the deployment tool’s application report. What’s required is that this scope is written in the procedure and that automatic patches can be tied to this class in reconciliation.

When should approval be obtained for an emergency change?

Before implementation, with a fast-track approval from an authorized person (the emergency change board or a defined authority). Full documentation and review are completed afterwards within the time set in the procedure. In the audit, both the trail of the prior approval and the date of the subsequent review are looked for.

How is segregation of duties achieved in a small IT team?

The basic rule is that the person who develops or implements a change doesn’t approve their own change. If the team is too small to achieve this, approval can be given by a manager outside IT, or an independent person periodically reviews the implemented changes from system records. This compensating control must be defined in the procedure and recorded.

Support

Contact us for support or a discovery session on this topic.

Contact form
[email protected]

Eretek Compliance

We work with you on compliance with ISO 27001, IATF 16949, TISAX and similar standards, covering risk assessment, control design, documentation and audit preparation.

Knowledge Center

Other articles