Meta’s delivery system is highly automated, so the strongest lever an advertiser controls is the quality of the signal it sends. That flips the usual audit order. On Meta, signal quality comes before campaign settings.
This guide assumes business portfolio ownership is already confirmed, as covered in Taking Over an Ad Account.
Signal: Is Meta Learning From Accurate Data?
Pixel and Conversions API Deduplication
When the same purchase arrives from both the Pixel and the Conversions API, Meta deduplicates using the event name and event ID. Missing event IDs mean one sale can count twice, and reported ROAS inflates. Events Manager shows whether each key event has both browser and server sources and how deduplication is performing.
Event Match Quality
Meta’s Event Match Quality documentation describes a score out of 10, rated Poor, OK, Good, or Great. It reflects how effectively server-sent customer information can match events to Meta accounts. Meta notes that better match quality can help advertisers see more conversions and lower cost per result. Low scores usually trace back to server events that send only IP address and user agent, without hashed email or phone number.
Multiple integrations. Inherited accounts often have a Shopify app, a tag manager server container, and an agency’s custom integration all sending the same events. Events Manager’s data sources view shows which partners are active, and duplicates should be reduced to one server-side path.
Structure: Is Fragmentation Blocking Learning?
An ad set exits learning after roughly 50 optimization events in the seven days after its last significant edit. An account producing 120 purchases a week across twelve ad sets gives none of them enough data. Most sit in learning or “Learning limited” indefinitely.
The fix is usually consolidation into fewer, broader ad sets. Mechanics are covered in Meta Ads Budget and Bid Strategies.
Campaign Budget and New Customer Audiences
Under Advantage+ campaign budget, check whether prospecting ad sets actually receive spend or whether retargeting absorbs all the budget. A report showing healthy CPA but no new customer growth may indicate that campaign-level budget is flowing entirely to retargeting. This check reveals whether the account is losing prospecting momentum despite seemingly good metrics.
Creative: Fatigue and Coverage
For the top-spending ads of the last 30 days, compare frequency with click-through rate over time. Rising frequency with falling CTR signals fatigue. Accounts running only static images also leave Reels and Stories inventory underused, because vertical video is missing.
Attribution: Are Periods Comparable?
In March 2026, Meta narrowed click-through attribution to link clicks only. Conversions after likes, shares, saves, and comments moved to a separate engage-through category. Period-over-period comparisons that span the change can show a false drop in click-through conversions, as explained in What Is the Attribution Window?. Mixed attribution settings across ad sets also make cross-campaign comparisons unreliable.
Billing and Country Fees
Since July 1, 2026, Meta ads shown in Turkey include a 5% location fee. This fee does not appear in the Ads Manager spend column; it shows separately on the invoice. When auditing, the ROAS shown in the panel differs from the true ROAS after fees are included. The calculation is covered in Turkey’s Country Fees on Ad Invoices.
Findings Mapped to Problems
| Finding | Likely problem |
|---|---|
| Reported purchases well above actual store orders | Missing Pixel and Conversions API deduplication |
| Event Match Quality at Poor or OK | Server events missing customer information parameters |
| Most ad sets in “Learning limited” | Account too fragmented for its conversion volume |
| Healthy CPA but few new customers | Campaign budget flowing to retargeting |
| Frequency up and CTR down | Creative fatigue |
| Click-through conversions down after March 2026 | Attribution definition change rather than performance loss |
Summary
Meta audits start with signal: deduplicated Pixel and Conversions API events and strong Event Match Quality. Structure comes next, because fragmented accounts never exit learning. Creative fatigue and format gaps follow. Attribution definitions and settings determine whether any before-and-after comparison can be trusted.