Scope: what this audit covers and what it does not
The scorecard covers a single Meta ad account and what feeds it: the pixel, any server-side event connection, the event set-up the account optimises towards, how campaigns and ad sets are arranged, how audiences and exclusions are built, how much creative variety each ad set has, whether that creative is wearing out, how budget is distributed and how results are measured and reported. It applies to accounts optimising for purchases, leads or reach, with the measurement rows adjusted to the objective.
It does not re-teach the work. Installing and testing events belongs to the Meta pixel set-up guide; building exclusions and resolving overlap belongs to the audience exclusions and overlap guide; the tick-box pass you run while fixing is the Meta Ads account audit checklist. It also does not judge the brand strategy, the offer or the website. Those questions sit with other scorecards in the audit library, notably the landing page and analytics audits, which often need running alongside this one.
Before you score: access and evidence
You need access to the ad account and to the events manager for the dataset it uses, plus admin visibility of the business portfolio so you can see who owns what. Fix an evidence windowThe fixed period of data an audit scores, agreed before scoring, long enough to cover a full cycle for the business and kept the same at every rescore. covering several creative cycles, because fatigue only shows over time, and keep that window for every rescore. If ownership turns out to sit with a former agency or a single person's login, the tutorial on setting up a Meta business portfolio is the reference for how that layer should look before the rest is scored.
Gather the event list with each event's source (browser, server or both), how duplicates are handled and the match quality the events manager reports. Export ad-level delivery with spend, reach, FrequencyAverage times a person saw an ad. and CPMCost per 1,000 impressions. by week, so trends are visible rather than averages. Note the audiences in use, including any Lookalike AudienceAudience similar to a seed list. seeds and their source lists, and the customer lists used for exclusions with their last refresh date. Finally, write down the business target for each campaign. Where awareness campaigns are judged on cost, the CPM calculator helps check that the target is consistent with the budget and reach the plan assumes.
Criteria and scoring
Rate each row missing, partial or solid. The gate rows decide the band on their own if they score missing.
| Criterion | What good looks like | Score guidance |
|---|---|---|
| Pixel events (gate) | Key events fire once, with correct parameters, on the right pages | Missing: absent or misfiring / Partial: fires but parameters wrong or incomplete / Solid: tested and correct |
| Server events and deduplication (gate) | Server-side events sent alongside browser events, with deduplication working | Missing: browser only with known loss / Partial: server events but duplicates / Solid: both sources, deduplicated |
| Optimisation event | Campaigns optimise for the deepest event with enough volume to learn | Missing: shallow event by default / Partial: right event, starved of volume / Solid: deliberate choice |
| Campaign structure | Few campaigns, each with a clear job; ad sets consolidated enough to exit learning | Missing: fragmented sprawl / Partial: some redundant splits / Solid: consolidated and purposeful |
| Audience approach | Broad, interest or lookalike choices made on purpose and documented | Missing: inherited with no rationale / Partial: mixed logic / Solid: deliberate and written down |
| Exclusions | Customers, recent converters and staff excluded where the offer requires it | Missing: none / Partial: lists stale or patchy / Solid: current lists applied consistently |
| Audience overlap | Prospecting and retargeting do not bid against each other | Missing: heavy overlap ignored / Partial: known, partly fixed / Solid: controlled |
| Creative depth | Each ad set has several genuinely different concepts and formats | Missing: one ad per set / Partial: variants of one idea / Solid: distinct concepts across formats |
| Placement and format fit | Creative made or adapted for vertical, feed and story placements | Missing: one crop everywhere / Partial: some adaptation / Solid: built per placement |
| Creative fatigue | Frequency and response tracked by ad; refresh before decline sets in | Missing: no tracking / Partial: refresh after results drop / Solid: planned rotation |
| Budget distribution | Budget concentrated where results justify it; learning not reset by constant edits | Missing: spread thin / Partial: some starved sets / Solid: concentrated and stable |
| Measurement and attribution | Attribution setting stated; results checked against analytics and the back office | Missing: platform figures only / Partial: occasional reconciliation / Solid: regular reconciliation |
| Testing practice | Tests isolate one variable and are logged | Missing: no tests / Partial: tests without records / Solid: logged test plan |
| Account hygiene | Ownership in the business portfolio, naming convention, no orphaned assets | Missing: personal ownership, no convention / Partial: mostly tidy / Solid: clean and owned |
Score the server events row against what the Meta Business Help Centre documents about deduplicationRemoving double counts when one conversion is reported by more than one source. and event matching, not against third-party claims. For fatigue, use the frequency and creative fatigue guide to decide what "before decline" means for the account's objective.
How to score and prioritise
Score the two signal gates first, then the optimisation event, because these three decide what the delivery system is learning. If either gate is missing, mark structure, audience and budget rows provisional: the pattern you see may be the system compensating for poor signal rather than a structural fault.
Then rank the remaining weaknesses by spend touched and by speed of effect. Exclusion and overlap fixes usually touch a large share of spend and work quickly, so they tend to rank high. Creative depth fixes need production time, so schedule them early even if they land later. Structural consolidation is valuable but resets learning, so time it away from peak trading periods. Anything touching a small campaign with stable results goes to the bottom regardless of how untidy it looks.
Reading the result: what to do with each band
Red means a signal gate scored missing. Hold budget increases and new campaign launches, repair the pixel or server events, and confirm deduplication in the events manager before trusting any result. Rescore once a full optimisation cycle has run on clean events.
Amber means the signal is at least partial but a large share of the other rows need work. Expect this to be the common outcome. Pick a short, ordered list, usually exclusions and overlap first, creative depth second, consolidation third, and change one area at a time so results stay readable.
Green means the signal is solid and most rows are too. Shift the effort to a creative testing rhythm and to the landing pages and offers behind the ads, and keep a quarterly rescore to catch drift.
From audit to action plan
Each partial or missing row becomes a plan line: the finding, the fix, the owner and the rescore date. Group lines by who does the work, because signal fixes sit with developers, exclusions with the media buyer and creative depth with the design team, and each group needs its own deadline. Present the band and the top few lines to the budget holder; keep the full table as the working copy.
A skincare brand with a red band
A direct-to-consumer skincare brand spends £9,000 a month on Meta. The audit finds browser and server purchase events both firing without a shared event ID, so the account reports roughly 1.6 purchases for every order in the shop's back office. Server events and deduplication score missing, which makes the band red. Creative depth also scores missing: eleven of fourteen ad sets run a single video. The plan fixes deduplication first and pauses a planned 30% budget increase. Three weeks later, reported purchases line up with orders within a few percent, the band moves to amber, and creative production for four new concepts is scheduled next.
Common mistakes
- Reading platform-reported results as truth. A broken signal inflates or hides conversions. Reconcile with the back office before scoring anything that depends on results.
- Marking broad targeting as a failure. The criterion is whether the approach is deliberate and controlled, not whether it uses interests.
- Counting ads instead of ideas. Five colour variants of one image is shallow creative. Score concepts.
- Fixing everything in one week. Simultaneous changes reset learning and blur cause and effect. Work the list in order.
- Ignoring ownership. An account owned by a former contractor's personal profile is a risk even when results are good. Score it and fix it.
- Scoring fatigue from averages. An account-wide average hides the one ad set burning out. Read frequency and response by ad, week by week.
