Print-ready version Back to the article
HMMNM Security
hmmnm.com

Governance & Compliance

ISO 27001 From Scratch Part 5: The Certification Audit — Stage 1, Stage 2 and What Fails

By Prabhu Kalyan Samal | Published 22 November 2026 | 9 min read

https://hmmnm.com/iso-27001-from-scratch-part-5-the-certification-audit-stage-1/

10 min read · 1,894 words

TL;DR: ISO 27001 certification happens in two audit stages: Stage 1 checks that your ISMS documentation exists and is coherent (often remotely, 1–2 days), and Stage 2 verifies it operates in practice through evidence sampling (typically 2–5 days on site). Most first-time failures aren’t technical — they’re traceability failures: risk acceptances with no review date, SoA exclusions with no justification, and backup plans that were never tested with an actual restore.

Authorization note: everything in this article is about your own lab ISMS — your own documentation, your own VMs, your own logs. Run every command against machines you own or have explicit written permission to audit. Nothing here involves attacking third-party systems.

Where You Are in the Journey

By Part 5, you’ve scoped the ISMS, run your risk assessment, written policies, built your Statement of Applicability (SoA), and operated the system for a while. In a lab environment, “operated” might mean 6–8 weeks of real activity: change records, access reviews, log reviews, an internal audit, and a management review. That operational history is not optional — certification bodies will not certify an ISMS that has never run.

The certification audit is where all of it gets tested against reality. Two stages, one goal: can an independent auditor conclude your ISMS conforms to ISO 27001:2022 and actually functions?

Choosing a Certification Body

Before the audit comes the paperwork. Your certification body (CB) must be accredited — in practice this means accredited by a national accreditation body (UKAS in the UK, ANAB in the US, DAkkS in Germany) under ISO/IEC 17021-1. An “ISO 27001 certificate” from an unaccredited CB may be worth very little to customers and regulators.

When shortlisting CBs, compare:

Sign the contract early. CBs book audits weeks out, and your certificate is dated from the Stage 2 close-out, not from when you started preparing.

Stage 1: The Readiness Review

Stage 1 is usually remote and lasts one to two days for a small ISMS. The auditor is asking one question: is this ISMS ready to be audited? They will not certify you at Stage 1, even if everything is perfect.

What the Auditor Reviews

Simulating Stage 1 on Your Own Lab

You can dry-run the document completeness check yourself. Keep your ISMS documents in a versioned folder and script the check:

$ ls -1 ~/isms-docs/
SoA-v1.3.xlsx
context-interested-parties.docx
infosec-policy.docx
internal-audit-2025-06.md
management-review-2025-07.md
risk-assessment-methodology.docx
risk-register-v2.1.xlsx
risk-treatment-plan-v1.2.xlsx
scope-statement-v1.1.docx

$ for f in scope-statement risk-assessment-methodology risk-register 
  risk-treatment-plan SoA infosec-policy internal-audit management-review; do
  ls ~/isms-docs/ | grep -qi "$f" && echo "OK  $f" || echo "MISSING  $f"
done
OK  scope-statement
OK  risk-assessment-methodology
OK  risk-register
OK  risk-treatment-plan
OK  SoA
OK  infosec-policy
OK  internal-audit
OK  management-review

Any MISSING line is a Stage 1 finding before the auditor ever sees it. If your SoA is a spreadsheet, also verify the risk register IDs it references actually exist:

# Check every risk ID referenced in the SoA exists in the risk register
$ cut -d, -f2 soa-controls.csv | grep -o 'R-[0-9]*' | sort -u > soa-risk-ids.txt
$ cut -d, -f1 risk-register.csv | sort -u > register-ids.txt
$ comm -23 soa-risk-ids.txt register-ids.txt
R-1042

comm outputting R-1042 means the SoA references a risk that isn’t in the register — exactly the kind of cross-document inconsistency auditors find in Stage 1.

Stage 2: The Implementation Audit

Stage 2 is the real audit. A typical small-scope audit runs 2–5 auditor-days, split between interviews, document sampling, and technical verification. The auditor samples — they cannot check everything, so they select a subset of controls, employees, and records and trace each one end to end.

Day-by-Day Shape of a Small Stage 2

Day Focus Typical activity
1 AM Opening meeting Audit plan agreed, scope confirmed, logistics set. Leadership interview: context, objectives, management commitment.
1 PM Clause walkthrough Leadership (5.1), roles (5.3), risks & opportunities (6.1), objectives (6.2), resources (7.1), competence (7.2), documented information (7.5).
2 AM Operations sampling Risk treatment evidence, SoA traceability, supplier management, change management, asset inventory.
2 PM Technical controls Access control, cryptography, logging, backup/restore, secure development — auditor may request live system evidence.
Final AM Closing meeting Findings presented: nonconformities, observations, opportunities for improvement. Certification decision follows CB review.

How Auditors Trace SoA Rows to Evidence

Every applicable SoA row must answer: is the control implemented, and where’s the evidence? A typical trace looks like:

  1. Auditor picks control A.5.15 (Access control) from your SoA, marked “Applicable — implemented.”
  2. “Show me the access control policy.” (Document)
  3. “Show me the user access review records.” (Record, with dates and sign-offs)
  4. “Pick a user — show me their entitlements match their role.” (System evidence)
  5. “Show me leavers’ accounts were disabled within your stated timeframe.” (HR exit list vs. account status)

Generate your own leaver-check evidence the same way. On your lab Linux host, list accounts and last login:

$ sudo lastlog | grep -v "Never" | head -5
Username         Port     From             Latest
root             pts/0    192.168.56.1     Fri Jul 18 09:14:22 +0000 2025
dev-jane         pts/1    192.168.56.1     Fri Jul 18 13:02:41 +0000 2025
ops-sam          pts/2    192.168.56.20    Thu Jul 17 10:31:08 +0000 2025

# Accounts never logged in and older than 90 days are review candidates
$ sudo lastlog -b 90 | grep -v "Never" 
Username         Port     From             Latest
old-contractor   pts/3    192.168.56.44    Tue Feb 04 15:20:11 +0000 2025

If old-contractor left the company in February and the account is still enabled, an auditor sampling leaver processes will find it in minutes — and it’s a classic nonconformity against A.5.18 (Access rights).

What Auditors Sample for Technical Controls

$ nmap --script ssl-enum-ciphers -p 443 192.168.56.10 | tail -20
| ssl-enum-ciphers:
|   TLSv1.2:
|     ciphers:
|       TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384 (ecdh_x25519) - A
|       TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256 (ecdh_x25519) - A
|   TLSv1.3:
|     ciphers:
|       TLS_AES_256_GCM_SHA384 (ecdh_x25519) - A
|_  least strength: A

If your cryptography policy says “TLS 1.2+ only, AEAD ciphers, cipher strength A or better,” this output is your evidence. If it shows TLS 1.0 enabled, the policy and reality disagree — traceability failure.

The Top Failure Patterns

After enough audits, the same failures recur. Fix these before your auditor arrives.

1. Risk Acceptances Without Expiry

Accepting a risk is legitimate — ISO 27001:2022 explicitly requires that risk owners approve acceptance and that acceptance be reviewed. The failure mode is a register full of “accepted” risks with no review date, no expiry, and no re-approval. An acceptance from 2023 that nobody has looked at since is effectively an untreated risk wearing a costume.

Fix: every accepted risk needs an owner, a rationale, an expiry date, and evidence of re-approval at the review. Audit your own register:

$ awk -F, 'NR>1 && $4=="accepted" {print $1, $6}' risk-register.csv
R-1007 2024-03-01
R-1019 2026-01-15
R-1042

$ awk -F, 'NR>1 && $4=="accepted" && ($6=="" || $6 < "2025-07-01") {print "STALE/UNDATED:", $1}' risk-register.csv
STALE/UNDATED: R-1007
STALE/UNDATED: R-1042

Two accepted risks need owner sign-off refreshed before Stage 2.

2. SoA “Not Applicable” Without Justification

In ISO 27001:2022, an N/A control in the SoA must include justification that it’s not applicable, and the justification must hold up. “We don’t have physical servers” fails if you have a rack in the office. Common bad examples:

Fix: for every N/A row, write one or two sentences explaining why the risk doesn’t exist within your scope, and reference the risk assessment conclusion that supports it. If you can’t write that sentence, the control isn’t N/A.

3. Backup Plans Without Tested Restores

This is the most common technical nonconformity. A backup policy and a backup job log prove backups run — they prove nothing about recoverability, which is what A.8.13 actually requires. If you’ve never restored, your backups are unverified data.

Test a restore in your lab and record it. Example with restic:

$ restic -r /backups/isms-lab snapshots --latest 5
ID        Time                 Host      Tags
8f2a91c3  2025-07-18 02:00:04  lab01     daily,etc
4b7d02ea  2025-07-17 02:00:03  lab01     daily,etc

$ restic -r /backups/isms-lab restore 8f2a91c3 --target /tmp/restore-test --include /etc
restoring  to /tmp/restore-test

$ diff -r /etc /tmp/restore-test/etc | head -5
Only in /etc: restore-test
Files /etc/fstab and /tmp/restore-test/etc/fstab differ
# diff output = your restore test evidence; attach it to the record with date + tester

Record: date, snapshot ID, what was restored, who verified, result. Keep these records — the auditor asks for the latest one.

4. Policies Nobody Follows

A 40-page password policy requiring 16-character rotation when your lab accounts use 8 characters with no expiry is worse than no policy. Auditors sample reality against documents, and mismatch goes either way: reality lagging policy or policy describing fantasy. Write policies you actually meet.

5. Internal Audit as a Checkbox

Stage 2 auditors read your internal audit report carefully. A two-page report saying “all fine, no findings” after one week of ISMS operation looks like theater. A credible internal audit has findings, however small, and evidence they fed into corrective action and management review.

Nonconformities: Major vs. Minor

Type Definition Example Consequence
Major Systemic failure or total absence of a required element No internal audit ever performed; SoA missing entirely; a mandatory control not implemented at all Blocks certification until corrected and re-audited (follow-up audit, extra cost)
Minor Isolated lapse that doesn’t undermine the ISMS One missed access review out of twelve; one record incomplete Doesn’t block certification; corrective action plan required, verified at next surveillance
Observation / OFI Not a nonconformity; improvement note “Consider automating log review” No action mandated

The Corrective Action Loop

Both major and minor nonconformities trigger the same discipline, mapped to clauses 10.1 and 10.2:

  1. Respond: contain and correct the specific instance (disable the stale account, run the missed restore test).
  2. Evaluate: was there a nonconformity, and what caused it? Root cause, not “human error.”
  3. Act: eliminate the cause (add automated leaver checks, add a restore test to the quarterly calendar).
  4. Verify: check the action worked.
  5. Update: change the ISMS documents if needed.

The CB typically gives you 90 days to close a minor nonconformity with evidence (root cause analysis + corrected records). Majors require a follow-up audit, on site or remote depending on severity. Certification is dated only after majors are cleared.

After You Pass

Your certificate is valid for three years with surveillance audits (smaller, sampled) at roughly years one and two, then a full recertification audit. The ISMS must keep operating between audits — auditors sample the intervening period, so gaps in reviews or records between certification

HMMNM Security · Independent security research and education · hmmnm.com
Printed from https://hmmnm.com/iso-27001-from-scratch-part-5-the-certification-audit-stage-1/
© 2026 HMMNM Security. For personal and internal team use. Please link to the original article when sharing.