How to Audit a Meta Ads Account in 60 Minutes

A 60-minute audit produces a prioritized diagnosis and the next evidence needed to act. It cannot establish a fixed amount of wasted spend or a causal sales lift. Review measurement and economics before changing creative or budgets.

Minutes 1–15: reconcile the business outcome

1. Choose paid orders or qualified new customers as the outcome. Record the allowable cost and whether it includes media alone or all acquisition costs.
2. Save dates, time zone, currency, attribution settings, order timestamps and the conversion-maturity cutoff. Count an order in its declared period; retain later conversions in the appropriate follow-up report.
3. Reconcile distinct order IDs, payment state, cancellations, refunds and existing versus new customers. Compare the same scope with platform-reported events.
4. Trace the conversion path and inspect event name, identity, value and currency. Multiple events for one order are a measurement question, not additional customers.

Use the event-identity checklist and break-even economics worksheet. Attribution and order timestamps answer different questions: a mature platform report is not automatically an accounting ledger.

Worked reconciliation: $600, 30 events and 15 orders

Hypothetical shop, not a customer audit: assume $600 media spend in a mature 14-day review scope, 30 platform-reported purchases and 15 distinct paid orders in the reconciled order cohort. Assume each retained order contributes $40 before media. Before investigating refunds, platform CPA is $600 / 30 = $20 while paid-order CPA is $600 / 15 = $40. Neither number is necessarily the cost of acquiring a new customer.

Keep the cohort and cost definitions consistent: do not divide company-wide acquisition spending by a selected campaign’s customers. Record the refund cutoff and distinguish gross orders from retained orders. If the two sources use different windows or attribution scopes, report the discrepancy as unresolved rather than declaring 15 duplicate events. Reconcile event identity before naming the cause.

Minutes 15–30: inspect allocation and delivery

List objective, optimization event, destination, country, economics and budget owner for each campaign. Investigate whether fragmentation prevents the current question from receiving enough evidence. Separation can be justified by different business outcomes; there is no universal campaign-count limit. Read the account’s current delivery status before proposing a budget change.

Minutes 30–45: reproduce the conversion path

Open the destination on a phone, follow the promised offer and submit valid input through payment or confirmation. Record device, browser, steps and expected behavior. In the hypothetical shop, a valid address fails on mobile and one order produces two event identities. These are assumed findings to demonstrate the repair log, not observed defects in NovaStorm or a customer account.

A repair is accepted when the recorded failure no longer reproduces and the expected outcome is verified. Mark its status as open, reproduced, fixed or independently checked; save the evidence and next review time. For creative hypotheses, use the fatigue diagnosis instead of treating every change as a defect. Once defects are resolved, use the creative testing framework to plan a separate message experiment with an approved outcome and loss allowance.

Minutes 45–60: audit units in the final report

Worked report example, not an observed product bug: a CTR stored as the fraction 0.012 should display as 1.2%. A value already expressed as 1.2 percentage points must not be multiplied by 100 again. Declare the source unit for each field before formatting a NovaStorm report or exported worksheet.

The aggregate is 20 / 5,000 × 100 = 0.4%, not the unweighted average of 1.2% and 0.2%, which is 0.7%. Check that the report contains every row in its declared account scope; label partial coverage. Record denominator, currency, period, time zone, gross/net revenue and platform events versus unique paid orders. Review whatever output is actually available in NovaStorm against the source; do not assume a missing aggregate or integration exists.

Finish with one owner and acceptance check for each high-priority action. A corrected percentage or repaired form is a technical result. Measure subsequent sales separately with mature, comparable cohorts and, for a causal claim, an appropriate controlled comparison. Prepare any proposed campaign change as paused campaign and local ad-set drafts until the evidence and loss allowance have been reviewed.