Knowing the audit date is rarely good news. In most organisations it marks the start of a period where a few people spend weeks collecting screenshots. Yet the information the audit wants already exists inside the organisation. The problem is not its absence, it is that turning it into evidence is done by hand.

Why Audit Preparation Turns Into a Project

An auditor usually works from a sample: they pick twenty changes that reached production last year and ask the same questions about each. Who requested it, under which rule did it go for approval, who approved it, what ran, was the executed text the approved text, and what was the outcome.

When those answers sit separately in the ticketing system, the pipeline, the database log and email, each sample becomes its own investigation. Twenty samples means twenty investigations. What is more, the resulting evidence pack is made of screenshots, and the auditor rightly asks when each image was taken and whether it could have changed afterwards.

Why a screenshot is weak evidence

A screenshot is a statement by whoever took it. When it was taken, which filters were applied and whether it was edited afterwards cannot be read from the image. An auditor will not reject it, but will mark it as weak evidence and ask for more samples. The weakness of the evidence is the main reason audits drag on.

What Belongs in an Evidence File

The evidence file for a request should tell that request's whole lifecycle in one place. To keep the auditor from having to ask follow up questions, these items belong in it.

  • The request identity and ticket number. This is where the link to the corporate change record is made.
  • The full text of the scripts. Not a summary, the actual text that ran.
  • The integrity status of each script. The digest computed at execution time is compared with the digest of the text today: unchanged, modified, or never executed.
  • The risk assessment and its rationale. Which rule set the band should be written down.
  • The rule set in force that day. Later policy changes must not blur a past decision.
  • The approval steps. Who completed each one, when, and with what note.
  • A segregation of duties summary. Who approved, who executed and whether the two overlap.
  • The execution record. On which server, when, who started it, the outcome and any error text.
  • Emergency details. If the request was declared an emergency: who declared it, whether it was reviewed afterwards and whether it was found justified.
  • The audit trail entries. A time ordered list of the events for this request, with the chain verified.

The file should be readable by a person and processable by a machine. A readable PDF lets the auditor file it away; a structured copy of the same data lets another system, or an independent audit tool, re evaluate the file on its own terms.

The Seal: Has the File Changed Since

The most critical property of an evidence file is not its content but the fact that its content can be verified. When the file is produced, a mathematical digest is computed from its contents. That digest is written both inside the file and into the audit trail at the same moment.

Verification works like this: the file is handed back to the system months later, the digest is recomputed and compared with the entry in the audit trail. If the two match, the file has not changed since the day it was produced. If they differ, the file is no longer evidence, and that in itself is information.

The strength of this chain depends on protecting the record being compared against. The audit trail should be signed with a secret key, the key should live on the configuration side rather than in the database, and a second copy of the trail should be kept in a separate database. Whoever reaches the database not reaching the key is what makes the chain impossible to reproduce.

The question the auditor really asks

An experienced auditor asks less about the content of the records and more about this: can the team that keeps them alter them? If the answer is that they could but would not, there is no control, only trust. A verifiable seal moves that question out of the realm of trust.

Exceptions Are Not Hidden

An evidence file is not a marketing document. A file where everything looks flawless raises suspicion rather than confidence. A healthy file records its own exceptions.

If the requester executed their own request, the file flags it as an exception. If a script was modified after execution, the integrity status shows it. A request declared an emergency but never reviewed appears with an empty review field.

Governance is not measured by the absence of violations. It is measured by whether violations surface. A visible exception can be managed; an invisible one, once found during an audit, puts the credibility of the whole process into question.

Producing Evidence Is Itself an Event

An evidence file contains the full detail of a request: the complete text of the scripts, which servers they ran on, who approved them. That is a sensitive package which can leave the organisation. The right to produce it should therefore be managed as a separate permission and not granted to everyone.

The second rule is that producing the file is itself written to the audit trail. Who exported an evidence file, for which request and when, must stay on the record. That same entry is where the seal is later compared against, so producing the file and making it verifiable are one and the same action.

Designing Your Own Evidence File

You can move towards this model without a product. Four principles are enough.

  1. One identity. Every item of work bound for production gets a number at birth, and that number travels through every stage.
  2. A digest at execution time. Prove what ran from a digest computed at the moment of execution rather than from the text afterwards.
  3. Freeze the rule set. Copy the rules in force onto the request when it is opened, so that changing the policy tomorrow leaves past decisions readable.
  4. Record the export. Record who produced the evidence and write the seal into that record.

With these four in place, audit preparation shrinks from a project to a button. Evidence stops being something prepared for the audit and becomes something produced by the work itself.

Frequently Asked Questions

Is an evidence file different from a report?

Yes. A report summarises a period and usually produces different results when run again. An evidence file belongs to a single request, freezes the state at the moment it was produced and carries its own seal. Reports are for management, evidence files are for audit.

Why should an auditor trust a file we produced ourselves?

What matters is not who produced the file but whether it can be verified. The auditor hands the file back to the system, the seal is recomputed and compared with the audit entry. The trail itself is signed and a second copy is kept in a separate database. The auditor treats the claim as a repeatable check rather than a statement.

Can evidence files be produced for our past requests?

The file contains whatever the record contains. If the rule set was never frozen in the past, that section stays empty; if no digest was computed at execution time, the integrity status appears as unverifiable. The file does not invent missing information, it states that it is missing. The date you adopt the model is therefore also the boundary of your evidence quality.

How long does preparing this file take?

Once the model is in place there is no preparation time; the file is produced from the request detail in a single action. The real time goes into establishing the model, and that is paid once, before the audit. What gets paid over and over, audit after audit, is the manual collection of evidence.

Look at a sample evidence file

Let us walk through the contents of a request's evidence file and how the seal is verified.

Book a Demo →