How to Audit Your Meta Ads Account: A 12-Point Checklist for 2026
A Meta ads audit checklist should test four things in order: measurement, structure, creative and waste. In 2026 that means 12 checks, starting with Event Match Quality (scored out of 10), Conversions API coverage and the March 2026 attribution change, then learning-phase health (about 50 optimization events in 7 days), creative diversity and frequency.
Most Meta accounts we look at are not failing because of bids or audiences. They are failing because the numbers feeding the algorithm are wrong, the structure keeps resetting learning, or every ad is a small variation of the same idea. This checklist is the order we work through an account, and it is built so you can run it yourself this week with nothing more than Ads Manager and Events Manager access.
It mirrors the format of our 10-step Google Ads account audit checklist, adjusted for how Meta actually works after Andromeda, Advantage+ defaults and the 2026 attribution rewrite.
Before you start: pull these four views
Set the date range to the last 28 days and open these before you change anything:
- Events Manager, Overview tab for each dataset: Event Match Quality per event, event coverage and deduplication.
- Ads Manager at ad set level with the Delivery column visible, so you can see Learning, Learning Limited and Active states.
- Ads Manager at ad level with frequency, CPM, CTR (link click-through rate) and cost per result.
- Your backend record of orders or qualified leads for the same 28 days, from Shopify, your CRM or a spreadsheet.
Write down three numbers before you begin: Meta-reported conversions, backend conversions and spend. Every finding below gets judged against those.
Phase 1: Measurement (checks 1 to 4)
Fix measurement first. Every other check depends on Meta optimizing toward the right events, and a bad signal makes good structure and good creative look broken.
Check 1: Event Match Quality on your money events
Meta scores Event Match Quality (EMQ) out of 10. According to Meta's Dataset Quality API documentation (updated June 2026), the score reflects which customer information parameters you send, how good that information is, and the share of event instances Meta can match to an account. Look at Purchase or Lead first, not PageView. If those score low, the usual cause is missing hashed email, phone or the click ID on server events.
Check 2: Conversions API coverage and deduplication
Pixel-only tracking is no longer a defensible setup. Meta's documentation reports event coverage as the 7-day average share of Pixel events also covered by the Conversions API with matching deduplication keys, and it flags coverage below 75%. Confirm that browser and server events share an event ID, otherwise you either double-count or Meta throws events away.
Check 3: Re-baseline for the March 2026 attribution change
On March 3, 2026, Meta announced that click-through attribution for website and in-store conversions would count only link clicks. Conversions that followed a like, save or share now sit in a separate engage-through category, and the engaged-view threshold for video dropped from 10 seconds to 5. Meta said 46% of Reels purchase conversions happen within the first two seconds of attention. Separately, PPC Land reported that the 7-day and 28-day view-through windows were removed from the Ads Insights API on January 12, 2026.
The audit question is simple: did your reported CPA jump in March without a matching drop in backend sales? If so, it is a reporting shift, not a performance collapse. Our breakdown of what the 2026 Meta attribution update means covers how to read the new columns.
Check 4: Reconcile Meta against your backend
Divide Meta-reported conversions by backend conversions for the same period. Neither number is "true", but the ratio tells you how much trust to place in in-platform CPA. If the ratio moves sharply month to month with no change in strategy, the problem is tracking, not media. For lead generation, reconcile against qualified leads, not form fills.
Phase 2: Structure and learning (checks 5 to 7)
Check 5: Learning phase health
Meta's guidance, as summarized by PPC Land in September 2026, is that an ad set generally needs about 50 optimization events within 7 days of its last significant edit to exit learning. Ad sets that cannot get there are labelled Learning Limited. Count what share of your spend sits in Learning or Learning Limited ad sets. If it is most of the budget, the account is too fragmented for its conversion volume.
Check 6: Edits that keep resetting learning
Significant edits restart learning: changing targeting, placements, the optimization event or bid strategy, adding a new ad, pausing for more than 7 days, or making large budget or bid changes. Open the account's activity history and count resets per ad set over the last 28 days. Weekly tinkering is one of the most common and least visible reasons Meta CPA never stabilizes.
Check 7: Advantage+ settings you did not choose
Advantage+ is now switched on by default when you create a Sales, App or Leads campaign, and Advantage+ Shopping has been renamed Advantage+ Sales, which allows multiple ad sets and up to 50 ads per ad set. That is often the right call: Meta reported roughly 10% lower cost per qualified lead in early testing of Advantage+ on Leads campaigns. But audit it as a decision, not a default. Check custom audience exclusions for existing customers, and check whether lead campaigns optimize for a qualified-lead event rather than raw form submissions. Our Advantage+ vs. manual campaign benchmarks show where each tends to win.
Phase 3: Creative (checks 8 to 10)
Check 8: Creative diversity, not creative volume
Andromeda, the retrieval system Meta described in December 2024, picks which few thousand ads out of tens of millions are even considered for a given person. Meta reported a 6% recall improvement and an 8% ads quality improvement on selected segments from it. The practical consequence is that targeting now happens largely through creative. Ten edits of one video read as one ad. Count distinct concepts per ad set: different hooks, formats, people, problems and offers.
Check 9: Frequency and fatigue
Sort ads by frequency over 7 and 28 days. For each high-frequency ad, check whether CTR has fallen and CPM has risen across the period. That pairing, not frequency on its own, is the signal that an audience is tired of the creative. Rotate new concepts in before the curve breaks, not after.
Check 10: Placement-fit creative
Check that you have vertical 9:16 assets for Reels and Stories, and that text and logos sit inside safe zones. Advantage+ placements will serve a square feed asset into Reels if that is all you give it, and it usually performs like it.
Phase 4: Audience and waste (checks 11 and 12)
Check 11: Overlap and self-competition
If you still run several manual ad sets, use the audience overlap tool in Ads Manager to see whether they target the same people. Overlapping ad sets bid against each other and split the conversion volume each one needs to exit learning. Consolidate before you add anything new.
Check 12: Spend that did no work
Look at the bottom 20% of ad sets and ads by cost per result. Anything that has spent several times your target CPA with no conversions is a cut. Then check placement and geography breakdowns for spend you never meant to buy. Our 6-step wasted ad spend audit walks through the same logic across every channel.
The NAMEX Meta Ads Audit Scorecard (2026)
Score each check pass or flag. The thresholds below are our working rules from auditing accounts, not Meta benchmarks, except where a Meta figure is named.
| # | Check | Where to look | Flag if |
|---|---|---|---|
| 1 | Event Match Quality | Events Manager | Purchase or Lead below 6 out of 10 |
| 2 | CAPI coverage and dedup | Events Manager | Coverage below 75% (Meta's flag) or no shared event ID |
| 3 | 2026 attribution re-baseline | Ads Manager columns | CPA jumped in March 2026 with flat backend sales |
| 4 | Backend reconciliation | CRM or store vs. Ads Manager | Ratio swings sharply month to month |
| 5 | Learning health | Delivery column | Most spend in Learning or Learning Limited |
| 6 | Learning resets | Activity history | More than one reset per ad set per week |
| 7 | Advantage+ defaults | Campaign settings | No customer exclusions, or optimizing to raw leads |
| 8 | Creative diversity | Ad level | Fewer than 3 distinct concepts per ad set |
| 9 | Fatigue | Frequency, CTR, CPM trend | CTR falling while CPM rises on top ads |
| 10 | Placement fit | Asset list | No 9:16 assets for Reels and Stories |
| 11 | Overlap | Audience overlap tool | Manual ad sets competing for the same people |
| 12 | Dead spend | Ad set and breakdown reports | Several times target CPA spent with zero results |
Fix flags in order. A flag in checks 1 to 4 outranks everything below it, because it means the other results cannot be trusted yet.
If your scorecard comes back with more than three flags, get a free Meta ads audit from our team and we will reconcile your Meta numbers against your backend and rank the fixes by expected impact.
Common Meta audit mistakes
Judging performance before re-baselining attribution. Comparing a post-March 2026 CPA to a 2025 CPA compares two different definitions of a click.
Auditing averages. An account-level CPA hides the two ad sets doing all the work and the six doing none. Audit at ad set and ad level.
Making every fix at once. Ten simultaneous edits reset learning everywhere and make it impossible to see what worked. Batch structural changes into one edit, then leave the account alone for a week.
Benchmarking against the wrong numbers. Use costs from your own market and platform. Our paid social benchmarks for Meta, TikTok and LinkedIn and our Q2 2026 CPC and CPM benchmarks by platform are a starting point, not a pass mark.
Frequently asked questions
How often should I audit my Meta ads account?
Run the full 12-point audit once a quarter and after any major Meta platform change, such as the March 2026 attribution update. Check learning status, frequency and creative fatigue weekly, because those three move fastest and cost the most when they drift.
What is a good Event Match Quality score on Meta?
Meta scores Event Match Quality out of 10 and does not publish an official target. As a working rule we treat anything below 6 on Purchase or Lead events as a fix-first item, and we push for the highest score the data you collect allows by sending hashed email, phone and click ID with each server event.
Why did my Meta conversions drop in 2026 without any change in sales?
Most likely the March 2026 attribution change. Click-through attribution now counts only link clicks, so conversions that followed a like, save or share moved into a separate engage-through bucket. Separately, the 7-day and 28-day view-through windows were removed from the Ads Insights API on January 12, 2026. Compare against your backend before cutting budget.
How many ads should each Meta ad set have in 2026?
Enough genuinely different concepts for Meta's retrieval system to match to different people, not minor variations of one idea. Advantage+ Sales campaigns allow up to 50 ads per ad set. In practice, most accounts we audit need more distinct concepts, not more ads, and a handful of truly different angles beats dozens of near-duplicates.
Can I audit my own Meta ads account or do I need an agency?
You can run every check in this list yourself with Ads Manager and Events Manager access. Where an outside audit helps is in reconciling Meta's reported conversions against your CRM or ecommerce data and deciding which fixes to do first. North American Media Experts offers that as a free audit.
Next steps
Run checks 1 to 4 today. They take under an hour and they decide whether the rest of your Meta data means anything. If you would rather have a second set of eyes, our paid social team and Toronto Meta ads specialists run this exact audit for brands across Canada and the US, with no minimum spend and no retainer fees.
Book a 20-minute intro call with Ryan or request your free paid media audit.