Meta Ads Audit Checklist 2026: Find Wasted Spend From Your Ads Manager Export

The short answer

A useful Meta Ads audit starts with the right export: campaign and ad set level, broken down by day, with spend, reach, frequency, link clicks, a named conversion event and its value. Then check tracking, cost per result against target, frequency, budget fragmentation and ad sets that spend without converting. The free PPC Audit Tool runs the measurable checks on that export in your browser.

Step 1: Export from Ads Manager correctly

In Ads Manager, choose the campaign or ad set level, set the date range (30 to 90 days), and use Breakdown → By time → Day. Then customise columns to include:

• Campaign name, Ad set name
• Amount spent, Impressions, Reach, Frequency
• Link clicks
• A named conversion event, such as Purchases or Leads, plus its conversion value

Export as CSV.

Why not the "Results" column?

"Results" counts whatever event each ad set is optimised for. If one ad set optimises for purchases and another for landing page views, "Results" adds them together as if they were the same thing. Your cost per result then means nothing across ad sets. The PPC Audit Tool warns you when it sees a Results column. Pick the named event your business actually cares about.

Why break down by day?

Without daily rows you can't see a pixel or Conversions API break. Meta exports without a time breakdown give one row per campaign for the whole period, and the audit will tell you the tracking-break check didn't run.

Step 2: Check tracking before anything else

Tracking problems show up in three ways:

• Conversions drop to zero for several days while spend continues. Usually a broken pixel, a Conversions API outage, a domain or consent change, or a site release.
• Conversions recorded but value is zero. Purchases are firing without the value parameter, so ROAS can't be measured and value optimisation has nothing to work with.
• Platform conversions far above your real orders. Usually a pixel and Conversions API sending the same event without deduplication (a shared event ID), or a very generous attribution setting.

The third one needs your order data to spot. Compare Meta-reported purchases with orders from your store for the same days. If Meta shows meaningfully more, check deduplication in Events Manager.

Step 3: Cost per result against target

Enter your target CPA and ROAS. The audit flags:

• Account CPA more than 15% above target, with an impact range.
• ROAS more than 15% below target. Remember Meta's default attribution includes view-through conversions, so Meta ROAS is usually higher than what your analytics shows for the same campaigns. Compare like with like.
• Campaigns at more than 2× the account CPA while taking at least 10% of spend. Before cutting, check the campaign's role. A prospecting campaign often has a higher CPA than retargeting, and cutting it can starve retargeting of new audiences later.

Step 4: Frequency and creative fatigue

Frequency is the average number of times each person saw your ads in the period. Rising frequency with rising cost per result is the classic sign of creative fatigue or an audience that's too small.

Average frequency over 30 days	Typical meaning
Under 2	Broad reach; rarely a fatigue problem
2–4	Normal for most prospecting
5+	Watch closely; common for retargeting and small audiences

The audit flags campaigns averaging 5 or more. Fixes: broaden the audience, rotate in new creative, or cap spend on small retargeting pools. Judge frequency alongside cost per result, not on its own; retargeting can run at high frequency profitably.

Step 5: Spend without conversions

Ad sets and campaigns that spent more than about twice your target CPA with zero conversions are flagged. Two exceptions to keep in mind:

• Awareness, reach and video view campaigns are not meant to convert directly. The tool recognises common names like "Awareness" and "Video Views" and marks them low severity.
• Long sales cycles. For high-value products, conversions can arrive weeks after the click. Check a longer date range before cutting.

Step 6: Structure and learning phase

Meta's delivery system needs enough optimisation events to learn. Meta's own guidance is roughly 50 optimisation events per ad set in a week after a significant edit. Accounts with many small ad sets often never leave the learning phase.

Signs of fragmentation in your export:

• Many ad sets each with a handful of conversions
• One campaign taking 60%+ of spend while dozens of others share the rest (the audit reports spend concentration)
• Similar audiences split across ad sets that compete with each other

Consolidating into fewer ad sets with more budget each usually helps more than fine-tuning targeting.

Step 7: Placements and settings worth a manual look

These aren't in a standard export, so check them in Ads Manager:

• Attribution setting per ad set (for example 7-day click and 1-day view). Keep it consistent if you compare ad sets.
• Placements. Meta's API changes in 2026 removed some placements for new campaigns. Check that older campaigns still deliver where you expect.
• Existing customer exclusions in prospecting and Advantage+ campaigns, if you don't want to pay to reach current customers.
• Special ad category settings for housing, employment, credit, social issues and politics.

A monthly Meta audit routine

When	Check	Time
Weekly	Tracking breaks, spend without conversions, frequency on retargeting	10 minutes
Monthly	CPA and ROAS against target by campaign, budget concentration, learning-phase ad sets	30 minutes
Quarterly	Attribution settings, audience overlap, placement performance, creative refresh plan	1–2 hours

Use the same export settings each time so results are comparable from one audit to the next.

What to send a client

Clients rarely need every finding. A useful update has three parts:

• What we found: the two or three findings with the biggest impact range, each with its evidence line.
• What we're doing: the change, its expected effect and when you'll check the result.
• What we can't see yet: the data gaps, such as no daily breakdown or no order data to reconcile with. Saying this upfront avoids awkward questions later.

The downloadable HTML report already lists findings by severity with their evidence and limits, so it works as an appendix.

Reading the report

Each finding includes severity, evidence (the file, the rows and the values), an impact range with its method, and a confidence score. Small samples, short date ranges, stale exports and missing columns all lower confidence, and the report lists every check it couldn't run. Download the HTML version to share with a client or teammate.

Going further

The open-source SpendSentry PPC agent for Claude Code builds on the same checks with AI analysis, test design and client reporting across 12 ad platforms. It can't change your account unless you approve the exact change, which matters because Meta's official tools now let AI agents create and edit campaigns.