What an approval audit proves
The approvals and compliance guide builds the approval matrix, defines what an audit trailThe record, attached to the work item, of what was reviewed, which version, who reviewed it, when, what they decided and the evidence relied on. holds and names three periodic audits in a sentence each. The approval audit is the first: it samples recent live work and checks that each item had the reviews the matrix required, with records. This tutorial runs it.
An approval audit proves two things. First, that the matrix is being followed: work that triggered a legal, accessibility or data protection row actually went to that reviewer. Second, that the records exist and point at the right version, so that when someone asks who approved a live ad, the answer can be shown rather than remembered. Both matter for the same reason. As the governance hub puts it, data protection law, including the GDPR, expects organisations to be able to demonstrate compliance, not just to comply; the site's GDPR entry summarises the regulation's principles. Claims are reviewed against the advertising codes the Advertising Standards Authority publishes in the UK, as the approvals guide states. An audit is how the team finds out whether its evidence would hold up, before anyone else asks.
The audit does not prove that the work was right. A complete approval record on an item that still contains a problem shows the process ran, not that the reviewer's judgement was sound. Where the auditor sees something that looks wrong in the work itself, it goes to the reviewer who owns that row, and the audit records only that it was referred.
The two sibling audits have their own pages. The access audit is the tutorial on reviewing who can access your marketing tools. The claims audit reviews the claims library for aged evidence and claims that no longer hold.
Before you start
- Have the current matrix and the one in force during the period. If the matrix changed, items are checked against the version that applied when they were approved.
- Get access to the records. The work system and the approval records on each item, as the work management guide describes them, plus the channel tools' version and change histories.
- Name an auditor who did not approve the work. Usually someone in marketing operations. Reviewers whose rows are checked see the findings before they are reported.
- Agree the grades in advance. The grading section below sets out a starting scheme; agree it with the reviewers so the results are not argued afresh.
- Know who owns data protection. If the audit finds work that used Personal DataAny information relating to an identified or identifiable living person, such as a name, email address, location data or an online identifier. without the review the matrix required, that person hears about it straight away.
- Read the five whys. The five whys framework is how each gap is traced to a cause in the process rather than a person.
The steps
Set the period and the scope
Choose the period the audit covers, usually the time since the last audit, and the work in scope: which channels, which markets, which teams and agencies. Write it down at the start, so the sample cannot drift towards whatever is easiest to check. Include work produced by agencies and regional teams if they publish under the brand, because gaps often sit where work crosses a team boundary.
Draw a sample of live work across channels and work types
List the work that went live in the period from the work system and the channels themselves, because the work system may not hold everything. Then sample across channels and work types: emails, ads, landing pages, social posts, printed material, content pieces. Weight the sample towards work the matrix treats as higher risk, such as items with claims, prices or new uses of data, and include some routine work too, because the fast lane needs checking as much as the full route. For an item that went through the fast laneA written, shorter approval route for low-risk, repeatable or urgent work, such as replies or posts from approved templates; anything risky leaves the lane., check that it met the lane's written entry rules as well as the launch check. How many items to take is the team's choice; read the results with the care the guide to reading small numbers recommends, treating one gap in a small sample as a lead rather than a rate.
Work out which matrix rows each item triggered
For each sampled item, look at the live version and list the matrix rows it triggered: any public-facing work needs brand review; a new claim, comparison or testimonial needs legal or compliance; a price or promotion needs the commercial owner as well; a new data source or audience needs the data protection lead; a new page or email template needs the accessibility check against the standard the organisation has adopted, such as the Web Content Accessibility GuidelinesThe W3C's Web Content Accessibility Guidelines, the widely used standard for making web content usable by people with disabilities.. Work this out from the item itself, not from what the intake form said, because the most important gaps are items whose intake answers missed a trigger.
Check each required review happened, by the right role
For each triggered row, find the review on the work item: who reviewed it, their role, when and what they decided. Check the reviewer held the role the matrix names, not a stand-in without authority, and that the decision was an approval rather than comments that were never closed. An approval given in email or chat counts only if it was copied onto the item, as the approvals guide requires; note where it was not, because those are approvals nobody could find later. Record each row as present, present but incomplete, or missing.
Confirm the approved version is the version that went live
Compare the version recorded in the approval with what actually ran. The creative approval SOP records sign-off with the version approved and releases only from the asset library, and it warns that editing approved copy directly in the ad platform is how unapproved claims end up live. Check the channel's version or change history for edits made after sign-off: changed copy, a new price, a swapped image. For social posts, the social media approval workflow keeps approval states in the calendar or scheduler, which is where to look. An approval on a different version is not an approval of what ran.
Grade each gap
Grade every gap found using the scheme agreed before the audit, so the report separates gaps that matter from paperwork. The grading section after these steps sets out a starting scheme. Grade the gap, not the work: an item can have a serious gap and still turn out to be fine, and the grade reflects the risk the gap created. Where a gap may mean an item broke a rule, refer it to the reviewer who owns that row and record the referral; do not decide it in the audit.
Trace each gap to its cause
For each significant gap, ask why it happened, using the five whys, until the answer is something in the process. Most causes fall into three groups. The matrix: a trigger that is missing, ambiguous or does not cover a new kind of work. The workflow: a gate that can be skipped, an intake question that does not surface the trigger, a channel where edits after sign-off are easy. Record-keeping: approvals given but never attached, or attached to the wrong version. When an answer names a person, ask why the process let one person's slip reach customers.
Fix the matrix, the workflow or the record-keeping
Agree a change for each cause, with an owner and a date. A missing trigger becomes a new row or a clearer description of an existing one. A skippable gate becomes a required step in the campaign workflow, such as the launch check refusing items without approval records. Edits after sign-off are closed by making a changed item go back through the steps the change affects. Missing records are fixed by making the work item the only place an approval counts. Where the gap involves content programmes, the content governance guide covers the editorial side.
Report the results and book the next audit
Report what was sampled, the share of items with complete approval records, the gaps by grade, the causes found and the fixes agreed. The approvals guide recommends reporting the share of work with complete approval records, review turnaround and return rates by gate alongside the team's operational measures. Report them as the health of the gates, never as a KPIA metric tied directly to an objective. for individual reviewers: as Goodhart's law warns, a reviewer judged on speed will review less carefully. Book the next audit, and when it runs, check first whether the fixes from this one are in place.
Grading gaps
Grades keep the report proportionate. Without them, a missing signature on a routine social post sits beside an unreviewed price claim, and both get the same attention. Adapt this starting scheme with the reviewers before the first audit.
| Grade | What it means | Examples | What happens |
|---|---|---|---|
| Serious | A required specialist review did not happen, or happened on a different version from what ran | A price promotion edited after legal approval; a new audience used without the data protection lead | Referred to the row's reviewer at once; cause traced; fix owned and dated |
| Significant | The review happened but cannot be shown, or was done by the wrong role | Legal approval in an email never attached; brand sign-off by a junior colleague | Cause traced; record-keeping or role fixed |
| Minor | The review and record exist but are incomplete | Approval with no date or no version label | Fixed in the record; tracked for patterns |
| Not a gap | The item was correctly routed, including correctly through the fast lane | A routine post from an approved template with a launch check | Counted as complete |
Track the grades over successive audits. A fall in serious gaps shows the process is improving; a rise in minor gaps can mean the records are getting sloppier before anything worse happens.
An online optician audits its approvals (illustrative)
Brivotte Optics audits its approvals
Brivotte Optics, an invented online optician selling glasses and contact lenses, ran its first approval audit on a quarter's work. The marketing operations manager sampled 60 live items: 20 emails, 25 ads and 15 landing pages, weighted towards items with prices, claims or new audiences.
For each item she listed the matrix rows it triggered from the live version, then checked the work items. 51 items had every required review on record, by the right role, on the version that went live. 5 had minor gaps, mostly approvals with no version label. 3 had significant gaps: legal approvals given by email and never attached.
One gap was serious. A price promotion email for contact lenses had been approved by legal with one price, but the version that went to customers showed a lower introductory price and different terms. The email platform's history showed the copy had been edited directly in the platform the day after sign-off, when the commercial team changed the offer. The auditor referred the item to legal, who reviewed the sent version and decided what to do about it.
The five whys traced the serious gap to a workflow cause: the email platform let anyone with editing rights change a scheduled email after sign-off, and nothing sent a changed email back to review. The team made scheduled promotional emails locked after approval, with any change creating a new version that returns to the reviewers the change affects. The email-attached approvals were fixed by making the work item the only place an approval counted, with the launch check refusing items without one. The report went to the head of marketing and the legal lead, and the next audit was booked for the following quarter, starting with a check that both fixes were in place.
Common mistakes
- Sampling only what is easy to find. Work outside the work system, from agencies or regional teams, is where gaps hide. Draw the sample from the live channels too.
- Taking the intake answers at face value. The gaps that matter most are items whose intake form missed a trigger; work out the rows from the live item.
- Accepting an approval on a different version. An approval of the draft is not an approval of what ran.
- Judging whether the item broke a rule. The audit checks the process; refer anything doubtful to the reviewer who owns the row.
- Treating every gap the same. Without grades, paperwork gaps bury the serious ones.
- Stopping at a person. "The coordinator forgot to attach it" changes nothing; ask why the process allowed an approval to live outside the work item.
- Turning results into reviewer targets. Review speed reported per person invites lighter reviews.
- Never checking the fixes. The next audit starts by confirming the last audit's changes are in place, or the same gaps return. The governance hub shows where this audit sits beside the matrix and the operating model.
