TL;DR
After auditing dozens of Meta ad accounts, the same agency mistakes appear again and again: reporting platform ROAS while the client bleeds cash, set-and-forget account management, months of creative neglect, targeting theater with 47 micro ad sets, scaling by budget slider alone, random testing with no framework, and hiding bad results behind jargon. None of these are sophisticated failures. They're all basic disciplines, skipped. If your agency does three or more of these, you're paying for a logo on your reports.
We've read enough audit reports to have opinions
There's a special feeling you get opening a new client's ad account for an audit. Within about twenty minutes, you can tell exactly what kind of agency was running it. Not from the results, from the shape of the neglect. The same seven mistakes show up so reliably we've started using them as a checklist. If an account has four or more, we already know the story before we look at a single metric.
This isn't a hit piece on agencies. We are one. It's a field guide for anyone paying one, and a mirror for anyone running one. Let's go.
Mistake 1: Reporting ROAS while the client bleeds cash
The classic. The monthly report shows a beautiful 4.2x ROAS in big friendly type. The client's actual revenue barely moved. How? Because platform ROAS and business reality are different numbers, and some agencies only report the pretty one.
Meta's attribution is generous by design. It takes credit for sales it influenced, sales it merely witnessed, and occasionally sales that had nothing to do with it. An agency that reports only Ads Manager ROAS, never blended ROAS or MER, is either innumerate or hoping you are. The honest version of the report has two numbers: what Meta claims, and what the bank says. The gap between them is the most important metric in the account, and it's the one most likely to be missing from your deck.
Ask your agency for blended numbers. If they get uncomfortable, you have your answer.
Mistake 2: Set-and-forget account management
You know the account. Launched in March, "optimized" in April, and then nothing changed for five months except the invoice. Same ad sets, same budgets, same three creatives slowly dying of old age while someone collects a retainer.
Meta's auction is not a crockpot. You can't set it and walk away. Audiences exhaust, creative fatigues, CPMs shift with seasonality, competitors enter your auctions. An account nobody touches for months isn't "stable," it's decaying, and the decay is gradual enough that nobody notices until the quarterly review, when everyone acts surprised.
Competent management is boring and regular: weekly creative rotation, budget adjustments based on signal, audience refreshes, quarterly structural reviews. If your agency can't show you a changelog of what they actually did last month, they didn't do anything.
Mistake 3: Creative neglect
Related to mistake 2, but big enough to deserve its own entry: agencies that treat creative as the client's problem, or as a one-time deliverable.
The numbers are unforgiving here. Creative is the dominant variable in Meta performance in 2026, and creative has a shelf life measured in impressions. An account spending real money with three ads from February is not being managed, it's being watered occasionally. We've audited accounts doing six figures a month with fewer live creatives than a lemonade stand has lemons.
The tell: ask when the last new concept launched, not the last new variant. Changing the headline color doesn't count. If the answer involves the word "quarterly," your creative pipeline is a creative puddle.
Mistake 4: Targeting theater
Open the ad account. Forty-seven ad sets. Micro-interests stacked like a Jenga tower. "Yoga moms 25-34," "yoga moms 35-44," "yoga moms who also like oat milk." It looks like precision. It's theater.
This structure made sense in 2018. In 2026, with Advantage+ audience and consolidated learning, it's actively harmful: fragmented budgets that never exit learning, ad sets competing against each other in the same auction, and reporting so granular nobody can read it. The complexity isn't serving the client. It's serving the agency's need to look busy.
Modern structure is almost boring: a few consolidated ad sets, broad audiences, budget concentrated where the signal is. If your agency's account structure needs a map and a legend, they're optimizing for billable complexity, not performance.
Mistake 5: Scaling by budget slider alone
Client: "We want to scale." Agency: moves the budget slider up. ROAS craters. Agency: "The algorithm needs time to adjust." Three weeks later: "We recommend scaling back to stabilize."
Scaling is a discipline involving creative supply, audience expansion, and staged budget increases. The slider is one tool in that discipline, and it's the last one you reach for, not the first. Agencies that scale exclusively by budget are telling you they don't have a creative pipeline, an expansion plan, or honestly much of a plan at all.
Ask what the scaling plan is besides the budget. If the answer is silence followed by "we'll monitor closely," you're the experiment.
Mistake 6: Random acts of testing
"We're always testing!" Great. What did you learn last month?
Testing without a framework is just spending money with extra steps. Real testing has hypotheses ("we believe UGC-style hooks will beat polished studio creative for cold audiences because our buyers distrust ads that look like ads"), isolation (one variable at a time), kill criteria decided in advance, and learnings documented somewhere other than someone's memory.
What we usually find instead: a new ad launched every few weeks because someone felt like it, no two tests comparable, winners and losers never written down, the same losing concept retested quarterly by different account managers who don't talk to each other. That's not a testing program. That's vibes with a budget.
Mistake 7: Hiding behind jargon when results are bad
Every industry has its version of this, but media buying has perfected it. "We're seeing some headwinds from auction dynamics and signal degradation, but our incrementality framework suggests underlying demand remains robust." Translation: results are bad and we don't know why.
Jargon has legitimate uses. "Frequency," "attribution window," and "learning limited" are precise terms for real things. The problem is jargon as fog: deployed specifically when the numbers are indefensible, to make "we don't know" sound like "we've got this." If your agency's explanations get more complex as performance gets worse, notice the correlation.
The agencies worth keeping explain bad months plainly: what happened, what they think caused it, what they're changing. Clarity when losing is the actual test. Anyone can sound smart when ROAS is 5x.
What competent looks like (the short version)
Blended numbers in every report. A visible changelog of weekly actions. New creative concepts launching regularly, not quarterly. Consolidated structure, not Jenga towers. Scaling plans with more than a slider. Tests with hypotheses and documented learnings. Plain-English explanations, especially when things go wrong.
That's it. None of this is exotic. It's just the job, done properly, every week. The bar is lower than you'd think, which is exactly why clearing it is a competitive advantage.
FAQ
How do I know if my agency is underperforming? Ask for blended ROAS vs platform ROAS, a log of last month's actual account changes, and the date of the last new creative concept. Vague answers to all three is your diagnosis.
What should a Meta ads agency report include? Spend, blended ROAS or MER, what changed in the account that month, what was tested and learned, and what's planned next. Anything less is a screenshot collection.
How often should my agency launch new creative? Enough to stay ahead of fatigue, which depends on spend. As a rule: if you can't name something new launched in the last two weeks, it's not enough.
Is a high ROAS in Ads Manager enough to judge my agency? No. Platform ROAS is one input. Judge on blended numbers against real revenue, and on whether the account is improving quarter over quarter.
Should I switch agencies or fix the relationship? If the problems are process-shaped (no changelog, no testing framework, vague reporting), that's fixable if they're willing. If the problems are honesty-shaped (vanity metrics, jargon fog), that's character, and character doesn't optimize.
What does a good agency onboarding look like? An audit before promises, honest talk about measurement and blended numbers, a creative pipeline plan from week one, and no guaranteed ROAS figures. Guarantees in advertising are a confession.
Written by the MetaMaxd team. We run Meta ads for brands that are tired of agency theater: no vanity metrics, no "hacks," just accounts engineered to make money. If this article felt uncomfortably familiar, our audits are the intervention.
