Six places money leaks out of a Meta ad account, and how to check each one.

Leak 1: budget sitting in ad sets that are below your own average

Start here because it is almost always the largest single pool.

Export the last 30 days at ad set level with spend, purchases, purchase value and ROAS. Compute the account's weighted average ROAS. Then flag every ad set that has spent more than a meaningful amount — enough to be statistically real for your volume, often around 20 to 30 times your target cost per purchase — while returning less than about 60% of the account average.

That flagged spend is your recoverable pool. It is not all waste, because some of those ad sets are doing prospecting work that pays off elsewhere. But it is the pool to interrogate, and in most accounts it is larger than the operator expected.

Leak 2: rising frequency on a shrinking audience

Frequency is the number of times the average person in your audience saw your ad. It rises when the audience is too small for the budget, and rising frequency reliably precedes rising costs.

Meta's own research indicates that performance typically begins to decline once frequency passes roughly 3 to 4 for direct-response campaigns, and practitioner data for ecommerce prospecting puts the useful band lower still, in the region of 1.8 to 2.5 over a rolling seven days.

The check: pull 7-day frequency by ad set. Anything above 3 on cold prospecting is burning money on repetition. The fix is either more creative, a broader audience, or less budget in that ad set — in roughly that order of preference.

Leak 3: audiences that overlap and bid against each other

Run three ad sets targeting overlapping interests and you are, in a real sense, competing with yourself in the auction. The symptom is a set of ad sets that each look mediocre while the campaign as a whole underdelivers.

This is a much smaller problem than it was five years ago, because broad targeting with the delivery system doing the work has largely replaced granular interest stacking. But it persists in accounts that were built in the old style and never rebuilt.

• Check for several ad sets in the same campaign with similar interest or lookalike definitions
• Check for retargeting audiences that are not excluded from prospecting ad sets — you are paying prospecting prices to reach people you already own
• Check that purchasers in the last 30 days are excluded from acquisition campaigns unless you are deliberately running a repeat-purchase play

Leak 4: creative that stopped working and nobody noticed

Ad-level reporting is where the truth lives, and it is the report most teams look at least.

Pull the last 30 days at ad level. For each ad with meaningful spend, compare the first seven days to the last seven. A click-through rate that has fallen 20% to 30% below its own baseline, with frequency rising, is a fatigued ad — and every further dollar in it costs more than the first dollar did.

The distinction that matters: a bad ad was never good, and should be cut quickly on a small budget. A fatigued ad was good and stopped being good, which means the concept may still work with a fresh execution. Treat them differently.

Leak 5: budgets that are technically constrained

Two mirror-image mistakes, both common, both invisible unless you look for them.

The first is an ad set hitting its daily budget ceiling early every day while performing above account average. That is a profitable ad set being throttled, and the waste is the opportunity you never captured.

The second is an ad set with a budget so large relative to its audience that delivery has to reach further and further into low-intent inventory to spend it. The fingerprint is frequency rising and cost per purchase climbing while the ad set never quite spends its budget.

Both are found the same way: compare budget, actual spend and delivery for every ad set, and look for the ones where the two do not agree.

Leak 6: campaigns that were paused, restarted and forgotten

Every account over a year old has them. A campaign built for a promotion that ended in March. A duplicate created for a test and never cleaned up. An ad set restarted during a busy week and never reviewed since.

Sort every active campaign by spend over the last 30 days and read down the list. Anything you cannot immediately explain the purpose of is a candidate. This check takes ten minutes and regularly finds more money than the sophisticated ones.

Making it a habit rather than a rescue mission

An audit like this recovers a chunk of budget once. A weekly rhythm keeps it recovered.

• Weekly, 20 minutes: frequency by ad set, ad-level click-through trend, anything below 60% of account average with real spend
• Monthly, an hour: the full six checks above, plus a read down the active campaign list
• Quarterly: tracking health audit, audience overlap, and a genuine look at whether the account structure still matches the business

The most expensive thing in a paid media account is not a bad decision. It is a reasonable decision that nobody revisited for four months.