Audit evidence. Question bank.
Questions auditors ask about sensitive data, and the evidence that answers them
Eight questions, each with the evidence item, its producer and date, a completeness check and the control behind it.
Co-founder and CEO · Updated November 8, 2026

On this page
Eight questions, five fields each
Auditors ask about sensitive data in eight ways, whichever framework they test against. For each question the table gives the evidence item, the system and date it comes from, a check that the item is complete and the control it supports.
Applies to audits against ISO/IEC 27001:2022, SOC 2 (2017 Trust Services Criteria, revised points of focus 2022), PCI DSS v4.0.1, the HIPAA Security Rule and the GDPR. Cite the version your auditor tests against, because control numbers change between editions.
The evidence owner fills the table and each system owner confirms the producer. A reviewer outside the team re-performs one item per section. Plan one working session per section: naming an owner for every producer takes longer than running the exports.
- Pick the rows your audit scope asks about
- Name the producer of each item: the query, report or scan
- Run it, then fingerprint the output with SHA-256 and date it
- Write the completeness check beside it and run that too
- Have a second person re-perform one item per section
The question bank
One row per question. The last column names the controls to cite.
| Question | Evidence item and producer | System and date | Completeness check | Controls |
|---|---|---|---|---|
| Which systems hold sensitive data? | Inventory, one row per system with its owner, from catalog queries and scans | Each source, with the scan or query date | Rows match the asset register, or each difference has a reason | ISO/IEC 27001 A.5.9; PCI DSS 12.5.2; GDPR Art. 30(1); HIPAA 164.308(a)(1)(ii)(A) |
| How is it classified? | Classification scheme, plus findings by data type from a scan or catalog query | Each source, with the scan date | A reviewer checks a sample of findings against the stored values | ISO/IEC 27001 A.5.12; SOC 2 C1.1 |
| Who can reach it? | Access export per source: users, roles, groups and service accounts, with the grants query | The source catalog, on the export date | A known account is listed; the row count is recorded | ISO/IEC 27001 A.5.15, A.8.2; SOC 2 CC6.1; HIPAA 164.312(a)(1) |
| Was that access reviewed, and acted on? | Review sheet: reviewer, decision per identity, reason, date and removal deadline | The review sheet, with the export it used | Every identity in the export has a decision; nobody certified their own access | ISO/IEC 27001 A.5.18; SOC 2 CC6.2, CC6.3; PCI DSS 7.2.4 |
| How is it stored? | Stored-form check on each sensitive column or file, and the encryption settings the source shows | Each source, with the scan date | The check covers every location the inventory marks sensitive | ISO/IEC 27001 A.8.24, A.8.11; GDPR Art. 32(1)(a); PCI DSS 3.5.1 |
| How long is it kept, and is deleted data gone? | Retention schedule beside the oldest record in each table; a re-scan after deletion | The table or file store, before and after | The re-scan repeats the first scan’s scope and method | ISO/IEC 27001 A.8.10; PCI DSS 3.2.1; GDPR Art. 5(1)(e), 30(1)(f); SOC 2 C1.2 |
| What did you not check, and why? | Coverage record: scanned, partial, failed, excluded or not connected, each with a reason | The scan log and the exclusion approvals | Every source in the inventory has a status | ISO/IEC 27001 clause 4.3; HIPAA 164.308(a)(1)(ii)(A) |
| Is it still true, and did the fix work? | Dated scans on a schedule; for each fix, the finding, the change record and a re-test | The scan history and the change system | The re-test repeats the original method, scope and account | ISO/IEC 27001 clauses 9.1, 10.2; GDPR Art. 32(1)(d); HIPAA 164.308(a)(8) |
Scope and gaps: the inventory is sampled
An auditor tests scope in two directions. They pick a system from your asset register and look for it in the inventory, then pick one from the inventory and look for it in the register. A system missing from either list is a finding before any data is read.
Annex A 5.9 asks for an inventory of information and other associated assets. PCI DSS 12.5.2 asks you to document the scope of account data and confirm it at least once every 12 months. Under GDPR Article 30(1), the record names categories of personal data, and the inventory is where you test those categories.
The coverage record answers the question about gaps. Clause 4.3 asks you to determine the scope of the management system, and a record of failed and excluded sources shows where that scope stops. Our guide to document scan exclusions covers the wording.
Rejected item: a spreadsheet of systems compiled from a questionnaire, with no date or method. It records what people believe. A coverage record that lists only successes fails the same way, because the auditor asks for the scan log and looks for the failures.
Classification and storage: values, not labels
Annex A 5.12 is where an ISO auditor tests your classification scheme and how it is applied. For SOC 2 reports that include confidentiality, criterion C1.1 asks the entity to identify and maintain confidential information. Findings by data type, at table, column or file level, are the evidence.
Show each finding with a sample a reviewer can check, masked where it holds personal data. The completeness check is the reviewer’s sample, recorded with the reviewer’s name and date.
Storage comes next. GDPR Article 32(1)(a) lists pseudonymization and encryption of personal data as example measures, and an auditor checks them against stored values. PCI DSS 3.5.1 requires stored primary account numbers to be unreadable. The evidence is a check of the values in each sensitive location, plus the settings each source exposes.
Rejected item: the phrase “encrypted at rest” from a policy or a cloud console, with no check of stored values. Disk encryption protects a stolen disk and leaves query results unchanged, so a check of the column’s values is the test that answers what it holds.
Access and review: export, decision, deadline
Start with the export. Annex A 5.15 and 8.2 cover access control and privileged access rights, and HIPAA 164.312(a)(1)(1)) requires access control that allows access only to those granted rights. The producer is the grants query, run against the source’s own catalog, with its text and row count kept beside the output.
The review is a separate record. For each identity the reviewer writes a decision, a reason and a date, and a removal deadline for anything revoked. PCI DSS 7.2.4 sets a review of user accounts and privileges at least once every six months. SOC 2 CC6.2 and CC6.3 cover authorizing and removing access.
Separate duties on the sheet itself. Nobody certifies their own access, and a reviewer other than the database administrator signs the administrator’s grants. Our database access review guide gives the queries.
Rejected item: a screenshot of an admin console with no query and no date. The auditor can’t run it again, and it shows one moment on one screen.
Retention and change: before and after
Retention is tested against what the systems hold. PCI DSS 3.2.1 asks for retention and disposal that limit storage to what is required. Annex A 8.10 covers information deletion, and Article 30(1)(f) asks for erasure time limits where possible. The evidence is the schedule next to the oldest record in each sensitive table.
For deletion, run the same scan before and after on the same scope. If the second scan reads fewer sources than the first, the comparison proves nothing, so keep both scan logs.
Change closes the loop. GDPR Article 32(1)(d) asks for a process for regularly testing the effectiveness of measures, HIPAA 164.308(a)(8)(8)) requires periodic evaluation, and clause 10.2 covers corrective action. For each fix keep the finding, the change record and a re-test, as set out in how to show an auditor that an exposure was closed.
Rejected item: a closed ticket with no re-test. It shows that work was requested. It doesn’t show what the system holds now.
Fingerprint and date each export
Run this on a read-only copy of the export. It appends the row count, a SHA-256 value and the UTC time to a manifest your reviewer can check later.
# PowerShell 7. Read-only on the export; the one write is the manifest line.
$file = '.\access-export-2026-11-08.csv'
$rows = (Import-Csv -Path $file).Count
$hash = (Get-FileHash -Path $file -Algorithm SHA256).Hash
$utc = (Get-Date).ToUniversalTime().ToString('yyyy-MM-ddTHH:mm:ssZ')
[pscustomobject]@{
File = Split-Path $file -Leaf
Rows = $rows
SHA256 = $hash
HashedAtUtc = $utc
RunBy = $env:USERNAME
} | Export-Csv -Path .\evidence-manifest.csv -Append -NoTypeInformation
Get-Content .\evidence-manifest.csv -Tail 1
# Expected: one CSV line holding the file name, the row count,
# a 64-character hash, the UTC time and the account that ran it.What this bank misses
The bank covers questions about the data itself. It leaves out the rest of each framework: policies, training, supplier management, physical security and the audit opinion. It also leaves out how auditors choose samples, which differs by firm and engagement.
A SHA-256 value shows a file hasn’t changed since you recorded it. It can’t show the export was complete when you made it, which is why each row carries its own completeness check. Your auditor decides what is sufficient, and this page is general information, not audit or legal advice.
For the data side of each framework, see the pages on ISO/IEC 27001, SOC 2 and PCI DSS.
IRIS produces evidence that supports your assessment, from the sources you connect, each item linked to its source and scan date. Capabilities vary by edition.
An inventory of sensitive data by source, with the scan date: IRIS produces part of this, in the sources you connect.
Which tables and files a scan couldn’t read, and why: IRIS produces part of this, for the tables and files the scan account can list.
Who can reach sensitive data, with a reviewer’s decision on each identity’s access to a source: IRIS produces part of this, where the source exposes permission data.
A check of stored values for masked, hashed, encrypted or plain form: IRIS produces part of this, on sampled values, once activated on your install.
Evidence items collected in a register you can export: IRIS produces this.
Whether the evidence is sufficient for a control: Not produced by IRIS.
A scored readiness view for the frameworks behind these questions: not in v1; ask us if you need one: Not produced by IRIS.
Explore audit evidenceRe-perform one row before fieldwork
Pick the row your last auditor tested hardest. Ask a colleague outside your team to run its producer from your written steps, then compare their row count and SHA-256 value with your manifest.
A match means the row is ready. A mismatch is your first finding, with a date on it. The evidence-readiness checklist shows what to file beside each row.
Official sources
- ISO/IEC 27001:2022: clauses 4.3, 9.1, 9.2 and 10.2; Annex A controls 5.9 to 8.24, ISO. Opened November 8, 2026.
- 2017 Trust Services Criteria, revised points of focus 2022: CC6.1 to CC6.3, C1.1, C1.2, AICPA. Opened November 8, 2026.
- PCI DSS v4.0.1: requirements 3.2.1, 3.5.1, 7.2.4 and 12.5.2, PCI Security Standards Council. Opened November 8, 2026.
- 45 CFR 164.308(a)(1)(ii)(A) and (a)(8): risk analysis and evaluation, eCFR, US Government. Opened November 8, 2026.
- 45 CFR 164.312(a)(1): access control, eCFR, US Government. Opened November 8, 2026.
- GDPR Article 30(1): contents of the record, EUR-Lex. Opened November 8, 2026.
- Regulation (EU) 2016/679, Article 32(1) and (2), EUR-Lex. Opened November 8, 2026.
Start with a free assessment of one database you choose.
Get a free assessmentQuestion bank questions
No. Each row lists its own controls, and items such as PCI DSS 12.5.2 apply only where that standard does. Drop the rows your audit scope doesn’t reach.
HIPAA’s documentation rule, 45 CFR 164.316(b)(2)(i), sets six years for the records it covers. For the other frameworks, agree a period with your auditor and write it into your evidence procedure.
No. Our recommendation is a second person for every review, because a reviewer who made the record can’t test it.
No. A spreadsheet records what people reported, and an auditor will want to re-run something against the systems. Pair it with catalog queries or scans so the work can be repeated.
Protect your brand. Start with your data.
Bring the question your last auditor asked, and see which rows of this bank an IRIS scan can fill from one database.


