Meta ads audit checklist: diagnose before changing your account

What should you save before the audit?

Save one baseline with the account, date range, currency, time zone, attribution settings and objective. Write the question you need to answer. An audit of expensive purchases requires different evidence from an audit of missing leads. Keep those scopes separate so the report produces a decision rather than a long list of unrelated metrics.

Export the original report before changing its columns or settings. Keep the comparison period with it and note promotions, stock outages, website releases and changes in the sales process. These are potential explanations to investigate, not corrections to apply automatically.

• Define success: a completed purchase, a qualified appointment or another business outcome.
• Assign an owner who can verify orders or CRM outcomes.
• Record what cannot be checked with the access available. Unknown is a useful result; a guessed pass is not.

How do you check access and measurement?

Confirm that the right people can inspect the right assets, then trace a real test outcome through the measurement system. A permission issue can hide evidence; a measurement issue can distort it. Neither should be treated as proof that advertising stopped working. Keep the test record free of customer identifiers when sharing it.

List the ad account, Page, Instagram account, data source and relevant integrations with their owners. Meta’s SDK distinguishes activity actors, applications and objects. Use that distinction to ask who can change an asset and which application performed a change, rather than attributing every edit to a person.

For measurement, compare event names, timestamps, values and currencies with the underlying order or lead. Inspect duplicate browser/server events and the selected reporting settings. Our Conversions API guide explains matching and deduplication. For a lead workflow, compare instant forms and website forms on accepted outcomes. Matching the number of attributed conversions to store orders exactly is not a valid universal test: the systems answer different questions.

Which profitability threshold should you use?

Use the contribution available before advertising to calculate an affordable acquisition cost. Revenue alone leaves out product cost, fulfillment, fees and other variable costs. Define whether refunds and repeat purchases are included. Otherwise two people can use the same word, ROAS, while making different decisions about the same campaign.

Hypothetical example: an order brings $100 in revenue and $60 in variable costs before advertising. Its contribution is $40. On that simplified basis, the break-even acquisition cost is $40 and the break-even ROAS is 100 ÷ 40 = 2.5. A target below the full contribution leaves room for profit and other costs.

That calculation does not include fixed overhead, tax, refunds or future repeat purchases unless you add them explicitly. Do not mix a first-order threshold with a lifetime-value estimate and call the result a measured profit. The ROAS calculator shows the formulas so you can check the assumptions.

How do you review structure and creative?

Ask what decision each campaign, ad set and ad supports. Separate entities when they need different controls or business economics, not simply to make a report look tidy. For creative, compare a consistent outcome and period, then identify the next distinct hypothesis to test. One disappointing day is not a complete diagnosis.

Read our campaign structure lesson before merging entities, and compare CBO and ABO before moving the allocation owner. Record budgets and targeting constraints so the comparison remains interpretable.

For ads, distinguish a message problem from a format problem, an offer problem and a destination problem. Check the landing page, availability and checkout on a phone. If a creative appears tired, compare it against its own earlier performance and use the creative fatigue guide. Do not declare a universal frequency threshold to be evidence for every audience.

How do you prioritize the findings?

Prioritize findings by confidence, business impact and reversibility. A verified broken purchase event needs a different response from a suspected weak hook. Keep an observation separate from its explanation, and put one proposed action beside it. A short list with owners is more useful than a score whose calculation nobody can inspect.

Our suggested categories are: fix a verified defect, investigate an unexplained change, test a hypothesis, or keep monitoring. For each item, preserve the evidence, state the uncertainty and set a review date. If several settings must change together, explicitly record that you will not be able to isolate their individual effects.

Read the activity history guide before undoing recent edits. If the account may be compromised, handle that as an access incident first. Meta recommends two-factor authentication and reviewing sessions; a marketing optimization is not a substitute for securing the account.

What should the final audit report contain?

Deliver the baseline, a findings table, a decision log and a review schedule. Include the raw evidence or a safe reference to it, and show what remains unverified. The reader should be able to reproduce a calculation, understand why a change was suggested and locate the person responsible for checking its result.

Use the free report template for consistent definitions. Keep account identifiers and customer data in private working files rather than a public case study. Redact them before asking a colleague or external reviewer to comment.

A useful review compares the agreed business outcome with the original baseline, using the same reporting settings. Record whether you adopted, reverted or extended the test. Adwize can help read results and propose changes through its optimization workflow; it does not make this checklist an independent assessment or remove your responsibility for the decision.