Meta Ads Audit Checklist: Step Zero Is Checking the Account Is Alive

1. Step zero: is the account alive?

Before any metric, pull the account list with the prompt below.

Watch for disabled, unsettled, pending risk review, in grace period and pending settlement. A disabled account comes with a reason, like being flagged for unusual activity with all ads paused.

Then check who has access. Owners taking an account back from an agency almost always find leftover users.

No agent? Export the last 14 days at ad and ad set level as CSVs and use the prompts below in ChatGPT. If you use a token, give it Analyst access, not Advertiser, so it can't spend.

# ask the connected agent, read only:
#   "list my ad accounts with account_status, is_queryable,
#    not_queryable_reason, has_payment_method and currency.
#    Quote any reason string verbatim."
#
# then, in Business Settings, read the people and system users
# with access to the ad account and remove anyone who left.

2. Check the tracking before you trust any number

Broken tracking makes every other finding wrong.

Is the server-side event arriving? Browser-only tracking loses events to ad blockers, so lower your confidence in everything after. If the page itself is the problem, fix it with the high converting landing page build.

Is it double counted? Events Manager has a deduplication tab for this. Browser and server both sending the same lead, without deduplication, gives Meta a number that doesn't exist to optimise toward.

Does the platform match the business? Put Meta's result count next to your CRM's for the same window. If Meta claims 80 leads and the CRM saw 41, that's your first finding.

Meta's event match quality score, from 0 to 10, rates how well server events match real people. Higher is better, and the trend matters most: a drop is a tracking change, not a creative problem.

3. Set the mode and the target

Decide sales mode or leads mode first. A clinic doesn't have a return on ad spend. It has a cost per booked consult.

Then work out your target: what a patient is worth over time, what you'd pay once to get one, and what share of leads become patients. That gives you a target cost per lead. Without it, the agent grades you on platform averages that aren't yours.

4. Build the columns, including 2 that don't exist

There are dozens of columns and you want about a dozen. Save a preset: spend, impressions, cost per 1,000, link click rate, frequency, results, cost per result, created date.

Then build 2 custom ones. Hook rate is 3-second video plays divided by impressions (the same thing as thumbstop rate). Hold rate is 15-second plays divided by 3-second plays. Published benchmarks use different denominators, so state yours.

Neither exists for image ads. A mostly-static clinic account gets graded on click rate and cost per result, and that's fine.

5. 2 numbers decide, the rest explain

Cost per result and return on ad spend decide. An ad that converts better beats an ad that gets more clicks.

A higher cost per 1,000 impressions often means a more valuable audience, because other advertisers want the same people.

6. Run the diagnostic tree

Pair 2 explaining metrics to name 1 cause. Each branch ends in a fix.

Good hook rate, weak click rate: people stop but don't click. Fix the offer or proof.

Weak hook rate: fix the first frame or first line, starting from what makes a good Facebook ad.

Good hook rate, weak hold rate: the part after the opening loses them, often because the next line doesn't deliver what the hook promised.

Lots of clicks, few link clicks: they're clicking the post, not the destination. Fix the copy and call to action.

Link clicks far above landing page views: the page is slow or broken, or taps are accidental. A small gap is normal. A gap of half or more is a bug.

Before you act: Meta may be holding back your best-looking ad on purpose. Raise its budget gently and watch.

7. Judge frequency by stage, and state the window

The most common wrong number in audits is one flat frequency limit. Frequency means nothing without a time window and a funnel stage.

On a 7-day window: cold audiences should sit low, and a steady climb is worth a look. Warm audiences tolerate more. Retargeting runs high by design, sometimes into double digits.

If you want a number, mine is a rule of thumb, not a Meta limit: on cold, a 7-day frequency above about 2.5 gets a look. It only means fatigue when click rate falls and cost per 1,000 rises with it.

8. Turn a finding into dollars

If you rank fixes by dollars, show the math. Take the ad set's spend and results for the window, get the actual cost per result, subtract your target, and multiply by the number of results. That's what the gap cost you.

Then sort by effort within the dollar ranking, so a 3-week creative project doesn't stall you: what ships today, this week, or is a project.

python3 - <<'EOF'
spend_7d, results_7d, target_cpl = 840.0, 12, 40.0
actual = spend_7d / results_7d
excess = (actual - target_cpl) * results_7d
print(f"actual ${actual:.2f} | over target by ${excess:.2f} per 7 days"
      f" | ${excess*52/7:.0f} per year if nothing changes")
EOF
# actual $70.00 | over target by $360.00 per 7 days | $2674 per year

9. Know the spend floor, then set 2 schedules

Per Meta, an ad set usually exits learning (Meta's early testing phase) after about 50 results in the week after its last significant edit, per ad set, not per ad. A clinic at $30 a day with 6 leads a week will never get a readable ad-level audit.

Below that floor, shrink the audit to 3 questions: is the tracking real, is the landing page working, and is there enough different creative running? Judge concepts, not single ads, and extend the window. Ad creative variations shows how to build enough different creative.

Meta's free opportunity score ranks recommendations by estimated impact, but it grades how well you follow Meta's recommendations, not your profit.

Every 60 to 90 days: the full audit. Weekly: the creative pass only: new ads, the kill list, 1 winner to feed. Call an ad fatigued only when frequency climbs, click rate falls and cost per 1,000 rises together. Act on the top 3 yourself, in Ads Manager, and check the activity history shows exactly what changed.

Check lead quality too. Ask whoever books consults for the first 3 questions they use to qualify someone, put them on the form, and accept fewer leads. 40 leads and 40 patients are different businesses.