/ Meta Ads audit

Meta Ads catalog audit example: what to keep and fix

Type: Anonymized client audit · Mode: Client sources, public version · Period: Recent 30 day window, exact dates withheld

This anonymized client example leads with decisions, not settings: what is working, where the data needs checking and which test should come next.

1 · fix1 · verify1 · test2 · protect

About this report and how to read the metrics

This version is based on a completed audit. Names, IDs, products, dates and commercial values were removed. Exact ROAS and CPA are withheld with their components.

Reading the metrics

A conversion is a goal action recorded by the advertising platform. Fractional values result from attribution. CPA is cost per conversion. ROAS is conversion value per unit of advertising spend, not profit. Conversion value depends on the account's measurement settings.

01 / in short

Summary

Current: product catalog · Other scenario: Direct leads →

Three decisions for the business owner

Start with the decision and next step. Sources, criteria and all 16 technical checks remain available in the appendix below.

The account does not need a rebuild. Keep the campaigns driving purchases, check measurement gaps and test new creative in a controlled way.

Keep: Current campaigns generate purchases. Do not split them into many small ad sets before there is enough data to support that move.
Fix: Check that each order is counted once and the product catalog is connected to the ads account. Otherwise campaign comparisons may mislead.
Test: Prepare three distinct creative approaches. Change one variable at a time and agree the budget and stop condition in advance.

All metrics

Metric	Value
audit window · exact dates withheld	30 days
active campaigns · structure reviewed	100%
purchase paths · deduplication checked	Pixel + CAPI
creative directions · for the next test	3

02 / audit map

What we checked.

Each tile is one area we checked. Select it to open the finding.

01 / Ads Manager and change historyProtectLow signal volume does not prove bad structureFragmentation would reduce signal per group. Apparently strong segments rely on too few purchases.02 / Events Manager, Pixel, CAPI and catalogFixPurchase is stable, but measurement is incompleteA working purchase event should not be disabled. Change one measurement variable at a time.03 / creative inventory and placementsTestThe creative system is too narrow for an audience verdictLarge individual orders are a plausible alternative to a stable audience effect.04 / Ads Manager and the limits of available business economicsProtectPlatform ROAS is not yet a profit thresholdHigher attributed revenue does not prove higher profit; that does not mean working campaigns should be paused.05 / Ads Manager change historyVerifyRecent edits weaken the ad comparisonThe difference could reflect delivery redistribution or learning, not just the ad message.

03 / where to start

First steps.

First, what every later decision depends on.

01Freeze baseline and close measurement gaps.→ Purchase is deduplicated, transaction IDs are present and any backend variance is explained.
02Run sequential creative tests.→ Each test changes one variable and continues only while its primary metric stays above the numeric stop threshold within the budget window.
03Scale only after economic proof.→ Margin, new customers and backend sales support the decision.

The full plan is in the detailed review below.

04 / findings

Every finding in three lines.

Numbers and evidence for your marketer sit under “details”.

Protect01 / Ads Manager and change history · confidence high

Low signal volume does not prove bad structure

What we see

Budget and current purchase cost produce only a few optimization events per week, while recent edits weaken period comparisons.

Why it matters

Fragmentation would reduce signal per group. Apparently strong segments rely on too few purchases.

What we do

The media buyer keeps consolidation. Acceptance: a new ad set stays separate only when purchase cost remains within the break-even threshold and beats the consolidated control; otherwise it returns to the consolidated structure.

details for the specialist

Details: Source: Ads Manager and change history. Period: 30 days. Status: observed. Confidence: high. Coverage: all active campaigns, ad sets and ads.

Fix02 / Events Manager, Pixel, CAPI and catalog · confidence medium

Purchase is stable, but measurement is incomplete

What we see

Purchase is deduplicated, while catalog linkage, domain evidence and upper event parameters are incomplete.

Why it matters

A working purchase event should not be disabled. Change one measurement variable at a time.

What we do

Analyst closes domain, catalog, consent and parameter gaps. Acceptance: test and real orders use one event ID, correct value and currency, with explained backend variance.

details for the specialist

Details: Source: Events Manager, Pixel, CAPI and catalog. Period: 30 days. Status: mixed. Confidence: medium. Coverage: all priority ecommerce events.

Test03 / creative inventory and placements · confidence high on asset count, medium on cause

The creative system is too narrow for an audience verdict

What we see

Delivery concentrates in one dynamic ad, while demographic differences rely on a handful of purchases.

Why it matters

Large individual orders are a plausible alternative to a stable audience effect.

What we do

The creative lead prepares three angles and the media buyer changes one variable per test. Acceptance: an angle scales only when its primary metric beats control while CPA or ROAS stays within the break-even threshold; otherwise the test stops.

details for the specialist

Details: Source: creative inventory and placements. Period: 30 days. Status: observed. Confidence: high on asset count, medium on cause. Coverage: all active ads.

Protect04 / Ads Manager and the limits of available business economics · confidence high on the risk, no claim about actual profit

Platform ROAS is not yet a profit threshold

What we see

The public extract lacks confirmed margin and a complete order register; a break-even threshold still needs agreement.

Why it matters

Higher attributed revenue does not prove higher profit; that does not mean working campaigns should be paused.

What we do

The owner supplies available category economics. Acceptance: a CPA or ROAS threshold is recorded before budget changes; unknown products are labelled separately.

details for the specialist

Details: Source: Ads Manager and the limits of available business economics. Window: 30 days. Status: data limitation. Confidence: high on the risk, no claim about actual profit.

Verify05 / Ads Manager change history · confidence medium on the cause of the difference

Recent edits weaken the ad comparison

What we see

Changes within the analysis window prevent a short before-and-after comparison from isolating one creative's effect.

Why it matters

The difference could reflect delivery redistribution or learning, not just the ad message.

What we do

The specialist keeps a control and changes one variable. Acceptance: window, budget, primary metric and stop rule are recorded before launch.

details for the specialist

Details: Source: Ads Manager change history. Window: 30 days. Status: observation. Confidence: medium on the cause of the difference.

Full findings table

Findings

Area	Finding	Evidence	Why it matters	What we do	Confidence	Decision	Details
Ads Manager and change history	Low signal volume does not prove bad structure	Budget and current purchase cost produce only a few optimization events per week, while recent edits weaken period comparisons.	Fragmentation would reduce signal per group. Apparently strong segments rely on too few purchases.	The media buyer keeps consolidation. Acceptance: a new ad set stays separate only when purchase cost remains within the break-even threshold and beats the consolidated control; otherwise it returns to the consolidated structure.	high	Protect	Source: Ads Manager and change history. Period: 30 days. Status: observed. Confidence: high. Coverage: all active campaigns, ad sets and ads.
Events Manager, Pixel, CAPI and catalog	Purchase is stable, but measurement is incomplete	Purchase is deduplicated, while catalog linkage, domain evidence and upper event parameters are incomplete.	A working purchase event should not be disabled. Change one measurement variable at a time.	Analyst closes domain, catalog, consent and parameter gaps. Acceptance: test and real orders use one event ID, correct value and currency, with explained backend variance.	medium	Fix	Source: Events Manager, Pixel, CAPI and catalog. Period: 30 days. Status: mixed. Confidence: medium. Coverage: all priority ecommerce events.
creative inventory and placements	The creative system is too narrow for an audience verdict	Delivery concentrates in one dynamic ad, while demographic differences rely on a handful of purchases.	Large individual orders are a plausible alternative to a stable audience effect.	The creative lead prepares three angles and the media buyer changes one variable per test. Acceptance: an angle scales only when its primary metric beats control while CPA or ROAS stays within the break-even threshold; otherwise the test stops.	high on asset count, medium on cause	Test	Source: creative inventory and placements. Period: 30 days. Status: observed. Confidence: high on asset count, medium on cause. Coverage: all active ads.
Ads Manager and the limits of available business economics	Platform ROAS is not yet a profit threshold	The public extract lacks confirmed margin and a complete order register; a break-even threshold still needs agreement.	Higher attributed revenue does not prove higher profit; that does not mean working campaigns should be paused.	The owner supplies available category economics. Acceptance: a CPA or ROAS threshold is recorded before budget changes; unknown products are labelled separately.	high on the risk, no claim about actual profit	Protect	Source: Ads Manager and the limits of available business economics. Window: 30 days. Status: data limitation. Confidence: high on the risk, no claim about actual profit.
Ads Manager change history	Recent edits weaken the ad comparison	Changes within the analysis window prevent a short before-and-after comparison from isolating one creative's effect.	The difference could reflect delivery redistribution or learning, not just the ad message.	The specialist keeps a control and changes one variable. Acceptance: window, budget, primary metric and stop rule are recorded before launch.	medium on the cause of the difference	Verify	Source: Ads Manager change history. Window: 30 days. Status: observation. Confidence: medium on the cause of the difference.

05 / limitations and scope

Limitations and scope

Limitations: Finance verified margin and a complete backend order ledger were unavailable, so platform ROAS is not called profit.

Coverage: All active campaigns, ads, events and available placement and demographic cuts were checked.

Sources:

Ads Manager and change history
Events Manager, Pixel and CAPI
Campaigns, ads and placements
Catalog and available order data

06 / detailed review

Full data for your marketer.

Every table and calculation from the audit. Owners can skip it.

Plan

What	Who	Result	Horizon
Freeze baseline and close measurement gaps.	Analyst and media buyer	Purchase is deduplicated, transaction IDs are present and any backend variance is explained.	7 days
Run sequential creative tests.	Creative lead	Each test changes one variable and continues only while its primary metric stays above the numeric stop threshold within the budget window.	30 days
Scale only after economic proof.	Marketing lead	Margin, new customers and backend sales support the decision.	90 days

Measurement

Status: Purchase arrives; quality can improve

Purchase arrives through browser and server paths with deduplication. Upper event parameters and catalog linkage still limit diagnosis.

What already works

Preserve purchase optimization and consolidation

Evidence: Pixel and CAPI deliver the primary event while the structure pools available signal.

Decision: Do not rebuild the account and tracking at the same time.

Target state and change rules

Target state: Stable purchase tracking, consolidated delivery, and a regular flow of genuinely different creative hypotheses.

How we test: Each test has a control, one variable, a budget cap, a primary metric, and a guardrail metric.

When we roll back: The new ad set or creative is stopped if break-even is breached or it fails to beat control within the agreed window.

Full audit map

16 checks

The three decisions above give an executive summary. The complete review register below shows the checks behind those decisions.

This is the check register from an anonymized client audit. Each check retains its conclusion and basis; private source links are removed from the public version.

Open the technical audit appendix

Measurement Integrity: the rebuild gate

The main event works, but catalog linkage and upper-funnel parameters are incomplete.

Check	Priority	Decision	What we established	What we established
MA-01	P0	Protect	Keep purchase as the primary optimization event.	Pixel and CAPI send purchase with deduplication.
MA-02	P1	Fix	Reconcile event ID, value, and currency.	One test order and one real order must pass without duplication.
MA-03	P1	Connect	Link the catalog to the correct dataset.	Incomplete linkage limits product diagnostics and dynamic ads.
MA-04	P2	Verify	Record consent and priority event parameters.	Signal quality must account for consent and data loss.

Architecture and signal volume

A small purchase volume does not prove the account should be fragmented.

Check	Priority	Decision	What we established	What we established
MA-05	P0	Protect	Keep ad sets consolidated.	Fragmentation would reduce events available to each optimization unit.
MA-06	P1	Define	Set the break-even threshold before changing budget.	Platform ROAS is not profit without margin and return data.
MA-07	P2	Verify	Separate learning effects from the underlying trend.	Recent edits weaken comparisons within a short window.
MA-08	P2	Keep consolidated	Do not create a segment from a few purchases.	A small number of large orders may explain the difference.

Placements, audiences, and delivery

Delivery is read with creative and volume rather than treated as an automatic instruction to switch something off.

Check	Priority	Decision	What we established	What we established
MA-09	P2	Monitor	Keep automatic placements as the control.	Volume is too low to attribute the difference to placement alone.
MA-10	P2	Verify	Compare age splits only after the event threshold.	A small demographic sample creates false precision.
MA-11	P2	Separate	Report prospecting apart from known customers.	New-customer share is required for a scaling decision.
MA-12	P3	Control	Read frequency beside creative fatigue.	Frequency without outcome trends does not prove saturation.

Creative system and test plan

Delivery is concentrated in one ad, so the next round needs distinct hypotheses rather than cosmetic variants.

Check	Priority	Decision	What we established	What we established
MA-13	P1	Test	Launch three distinct creative approaches.	The current set cannot separate audience strength from one ad's strength.
MA-14	P1	Change one variable	Do not change hook, format, and audience in one test.	Otherwise the team cannot identify the cause of the result.
MA-15	P2	Stop by rule	Set the budget and numeric stop threshold before launch.	A rule chosen after seeing results biases the test.
MA-16	P2	Scale	Move a winner only on primary and guardrail metrics.	The creative must improve the result without breaching CPA or ROAS limits.

Method and appendices

Pixel / CAPI event and deduplication map
Campaign, ad set, and protected-setting register
Creative hypothesis and placement matrix
Test, stop, and scaling protocol

What the client gets

Structure and budget decision
Pixel, CAPI and catalog review
Creative test map
Protected settings register
Implementation and recheck plan

/ next step

Next step with UPLIFY

Want the same clarity for your Meta Ads account?

See what to keep, fix and test first. The audit does not change anything in your account.

Discuss an audit for your business