How to Evaluate a Meta Ads Agency Without Getting Burned

Evaluating a Meta ads agency is a diagnostic exercise, not a vendor selection. Meta in 2026 rewards creative diversity, fast feedback loops, and honest measurement. The agency you hire has to produce on all three.

The questions that separate a real practitioner from a rehearsed pitch are specific. What does the creative testing framework look like on a live account. What is the hit rate on tested creative. Who actually runs the account on a random Tuesday morning. How does the agency reconcile platform return on ad spend (ROAS) with business revenue. Who owns the Business Manager (Meta’s account-administration interface where ad accounts, pixels, and pages are owned and assigned) and the pixel. Order matters. Creative and measurement come first, because every downstream claim depends on them.

Meta cost per mille (CPM) has moved 30 to 40% year over year from 2024 to 2025 across $30M+ a month managed across six channels, which means every wasted impression is more expensive than it was two years ago. Which in turn makes Meta ads management more important to get right. The questions below are written from inside that managed book, with a large Meta share inside that total. They apply to every Meta ads agency you will evaluate, including the ones who write articles like this one. Use them.

Meta evaluation is hard in 2026 because the platform’s performance model has shifted faster than most agencies have updated their playbooks. Creative diversity drives results more than interest targeting. Platform reporting overstates contribution because of iOS attribution gaps and view-through credit. Most agency websites still list the same five selling points: senior team, custom strategy, transparent reporting, creative capability, proactive communication. Those claims are simply not specific enough to separate one agency from the next by the time you reach a shortlist.

The asymmetry of the conversation works against the buyer. The agency has run dozens of pitch calls in the last quarter alone. The buyer has usually run fewer than five total agency evaluations across an entire career, often after getting burned twice by a previous agency. The result is a meeting where the agency has a script and the buyer does not. The questions below are designed to force a demonstration instead of a description.

Quick takeaways:

• Every agency website now claims creative, strategy, and data. By shortlist stage those claims are simply not specific enough; the buyer needs diagnostic criteria, not category claims.
• A rehearsed pitch sounds hedged and abstract: “usually,” “typically,” “depending on the account,” “data-driven,” “always testing.”
• A diagnostic answer sounds specific: named people, numeric benchmarks, Meta-specific mechanics, willingness to say what is currently broken in your account.
• Meta CPMs are up 30 to 40% year over year from 2024 to 2025 across $30M+ a month managed across six channels, which raises the cost of every wasted impression. The questions below filter for waste.

A senior paid media team should be able to answer every question with a number, a named person, or a specific change. Vagueness anywhere in this list is itself a signal.

Ask these ten questions in order. Creative and measurement come first because Meta performance lives there in 2026. Team and process come second. Contract and ownership come third. Each question has a diagnostic answer and a rehearsed pitch answer, and the gap between the two is what tells you who can run the account successfully.

1. Walk me through your creative testing framework on a live account.
2. What is your hit rate on tested creative over 30 and 90 days?
3. How do you think about creator and whitelisted ads, advertorials, and brand-page creative? When do you use each?
4. How do you reconcile platform ROAS with my actual revenue in the customer relationship management (CRM) system or profit and loss (P&L) statement?
5. Who is actually managing my account day to day, and how many other accounts do they manage?
6. What is your 12-month client retention rate?
7. What is your free audit going to find, and will you name dollar impact per line item?
8. How do you handle Advantage+ Shopping Campaigns and Advantage+ Audience on retargeting versus prospecting?
9. Who owns the Business Manager, the pixel, the Conversions API (CAPI) server, and the custom audiences?
10. What is one specific thing you would change in my Meta account in the first 30 days, and why?

Each question is unpacked in its own H2 below, with what a diagnostic answer sounds like and how a rehearsed pitch falls short.

Ask the agency to walk through a testing framework with real numbers, not a deck. There are four reasonable substrates for the answer: a live account they own, a sanitized case from a current client, a recent self-run test on their own brand account, or a dry-run plan against your own account on a whiteboard. An agency that refuses all four is hiding the framework.

A good answer has named formats, named personas, a specific creative volume per week, a budget split between proven winners and new tests, and a named rule for when to kill an ad. A rehearsed pitch answer leans on words like “data-driven,” “iterative,” and “always testing,” without a single number. Meta’s 2026 thesis is that creative diversity, not interest targeting, is the primary algorithmic signal on Facebook and Instagram. An agency that does not talk about creative diversity the way a paid search agency talks about match types is not ready for Meta.

Quick takeaways:

• A good answer includes weekly creative volume at a specific number, a formal budget split (80% winners and 20% tests is a common shape), creative diversity expressed as different angles and formats rather than hook variations on the same angle (an angle is a strategic claim like “fastest-shipping coffee on the market”; a hook is the opening line that delivers it, like “I tried four coffee subscriptions and only one shipped in two days”), a named kill criterion (cost per acquisition (CPA) multiplier or significance threshold), and a named process for scaling winners through post ID preservation (using the same Facebook post ID across ad sets so the social proof, comments, likes, shares, accumulates on a single ad rather than splintering).
• A rehearsed pitch answer hides behind tools lists and “we test regularly.”
• Confidentiality is a real constraint, but it is not a defense for refusing every substrate. A real framework is portable. The agency can walk through their own brand account or build the plan against your account on a whiteboard.
• One audited direct-to-consumer (DTC) account ran 96% of spend on video while statics had the lowest CPAs. The pattern, creative concentrating on one format while the data points elsewhere, repeats across 200+ audited accounts. An agency that cannot read that pattern is shipping on autopilot.

Ask for the number. Hit rate on Meta is the percentage of tested ads that reach a winning threshold over a defined window. A working program produces 5 to 7% winners (based on 250 to 612 ads tested across 200+ audited accounts) out of all ads tested over a 30 to 90 day window at scale. If an agency cannot give you a hit rate from an active account, they are not running creative testing as a structured motion (a repeatable, measured process), they are shipping ads and hoping. That is the structural problem.

Define the terms before asking, so the agency cannot dilute the answer. A winner is an ad that hits the account’s target cost per acquisition (CPA) or beats the rolling 90-day account CPA (CPA averaged over the trailing 90 days at the account level, rather than per-campaign or per-ad-set) by 20% or more, sustained over at least 7 days after Meta’s learning phase exits, at meaningful spend (typically $500 to $2,000 in test budget) and 1 to 2 conversions per day. Loose definitions, where any ad that ever beat CPA for one day counts, inflate the number into uselessness.

Quick takeaways:

• A good answer: “On a similar-sized account over the last 90 days we tested 250 to 400 ads, 5 to 7% reached our winning threshold, and here is what we define as winning.”
• A rehearsed pitch answer: “We test until we find winners,” or “Every account is different.”
• Creative as a structured motion means defined weekly volume, defined formats, defined kill rules, and a defined hit-rate review. Not “always testing.”
• Hook variations on the same angle do not count as different ads for hit-rate purposes. A 1 to 2% hit rate across a full quarter is a failing program. A 10% or higher hit rate usually means cherry-picking or a loose definition of winner.

Creator and whitelisted ads, advertorials, and brand-page creative are the three Meta formats with the widest performance spread in 2026. The agency’s answer tells you whether they have run creative on scaled Meta accounts. Creator and whitelisted ads run from a creator’s own handle typically outperform brand-page creative 2 to 3x on CPA. Advertorial landing pages cut CPAs 30 to 40% on DTC subscription offers by warming cold traffic. Brand-page creative is the default and usually the worst performer at scale.

A good answer engages with your account specifically. “Your account is almost all brand-page creative. We would run whitelisted creator ads through three creator handles for the top-funnel personas, and route advertorial landing pages for your highest-lifetime-value (LTV) product. Based on similar accounts, that is 2 to 3x lower CPA on the creator side and 30 to 40% lower CPA on the advertorial side.” A rehearsed pitch answer falls back on “we have creator relationships” or “we do UGC.”

Quick takeaways:

• The difference between user-generated content (UGC) and whitelisted creator ads is large. UGC is creative footage. Whitelisted is running that footage from the creator’s own handle with their social proof attached. The CPA advantage comes from whitelisting, not from the footage itself.
• One direct-to-consumer (DTC) brand audited by Opascope ran creator-handle ads at a mid-thirties CPA against $90 from brand-page creative on the same offer. Same creative footage, different surface.
• Advertorial landing pages convert better than category landers because they warm cold traffic with story before the offer. An agency that builds advertorials in advertorial landing pages is buying CPA reductions through the landing page rather than only through the ad.
• Creative on Meta is built and bought in service of the paid program. It is not a standalone deliverable.

Meta’s reported ROAS is structurally inflated. iOS attribution gaps (the conversions Meta cannot see because iOS Apple Tracking Transparency suppressed the pixel signal), view-through credit (the post-impression conversions Meta counts toward ROAS even when no click happened), and platform self-reporting (Meta scoring its own ad performance) all push the number up. A good Meta ads agency reconciles platform ROAS to a blended ROAS that uses total revenue over total ad spend, and names the gap between the two in a standard report. A rehearsed pitch agency reports platform ROAS as if it is revenue.

A good answer: “Platform ROAS on Meta usually overstates by 50% or more on the accounts we manage. We reconcile monthly against blended ROAS and against your CRM for lead-gen accounts. Here is what that looks like in our weekly report.” A rehearsed pitch answer: “We report on ROAS from the platform.” For business-to-business (B2B) accounts, the reconciliation is to qualified leads and closed-won revenue, not to platform-defined “leads” events. For a senior leader who owns the paid social budget without sitting in Business Manager day to day, this is the single most important question on the list. If you cannot trust the ROAS number, you cannot trust the budget, which means you cannot defend the spend to the board.

Quick takeaways:

• Across 200+ audited accounts, roughly half show underreporting of revenue against the platform number, with the gap reaching 50% or more in many of them. The pattern is almost always structural, not a settings fix.
• One B2B SaaS company moved cost per qualified lead from $512 to $103, an 80% reduction, after rebuilding paid social and the attribution layer in tandem. The CPA on the platform looked worse for the first month while CRM-true cost per qualified customer dropped.
• A weekly report that does not show a blended ROAS line item is reporting half the picture.

Ask for named people and an account ratio. The pitch partner is almost never the day-to-day account manager. An account-to-strategist ratio above 8:1 means the person running your Meta account has fewer than 20 hours a month to spend on it, which is not enough at scale. The whole question is who is in your Business Manager on a day-to-day basis, not who is on the pitch call.

A good answer is specific: named people, LinkedIn profiles, an explicit account ratio of 8:1 or lower, and an explicit senior escalation path with a named senior strategist or director, a weekly or bi-weekly account review cadence, and a service-level agreement (SLA) for escalation response. A rehearsed pitch answer reaches for “our senior team,” “you will have dedicated specialists,” or “we have a strong bench.” Opascope caps senior Meta strategists at 8 accounts. The math is simple: a strategist carrying 10 or more accounts has fewer than 20 hours per month for any single account, which is not enough at scale.

Quick takeaways:

• Ask for the day-to-day strategist by name, not just by role.
• Ask for the strategist’s most recent calendar week on an account of similar size: how many hours, what they did.
• Watch the seniority gap between pitch partner and day-to-day strategist. A senior pitch partner pairing with a junior account lead is the most common quality erosion in agency relationships.

Retention is the strongest trailing indicator of agency quality. Ask for a number, and ask explicitly for the 12-month window so the comparison is honest (any agency stretched across 10 or 20 years can quote a low headline retention rate that hides recent churn). Above 90% over 12 months is exceptional. Between 75 and 90% over 12 months is acceptable. Below 75% over the same window is a warning. An agency that refuses to share a 12-month retention number, or hides behind “most clients stay multiple years,” is signaling something about its own data. Every agency has a churn rate that is meaningfully above zero, so the willingness to volunteer the number is the real test.

A good answer is a specific percentage paired with a willingness to describe churn patterns. The follow-up that separates the two: “What is your churn pattern? When do clients usually leave, and why?” The diagnostic agency tells you. The pitch agency does not. For a C-level leader reading this (CFO, CEO, CMO, or CRO who owns the paid line), the 12-month retention number translates directly: an agency at 60% is going to spend the first half of every year onboarding new clients, which is a cost the existing accounts pay for in attention.

The free Meta audit is the most useful part of the evaluation, if you know how to grade it. A diagnostic audit names specific findings with a dollar impact per line, names problems the buyer did not already know about, and tells the buyer which findings they can fix without hiring the agency. A rehearsed pitch audit lists generic observations, recommends a retainer as the only solution, and avoids numbers.

Grade the audit on three axes. Specificity: does it name your actual campaigns, ad sets, creatives, conversion actions, and audiences. Dollar impact: does each finding carry a numeric estimate of wasted spend, missed revenue, or conversion-rate gap. Honesty: does the agency tell you which findings you can act on without hiring them. The willingness to do that last one is the strongest single quality signal in the evaluation, because it forces the agency to risk sharing information that may not lead to a paid engagement.

Quick takeaways:

• The buyer’s test: hand the audit to a strong internal marketer. Can they act on it without needing the agency at all.
• Across 200+ audited accounts, 50%+ of accounts are underreporting Meta revenue because of attribution setup, not creative or targeting. If the audit skips tracking, it is a rehearsed pitch audit and you can ignore the rest of it.
• Audit findings without dollar amounts attached are not findings. They are observations. The point of the audit is to give the buyer a number to act on.
• For a CFO, CEO, CMO, or CRO who owns the paid line: a free audit that names dollar impact lets you cost-justify the engagement before you sign. An audit that does not name dollar impact gives you nothing to defend on the budget side.

For a longer view of the patterns these audits surface, see common audit findings.

Advantage+ (Meta’s auction engine for creative diversity, powered by the Andromeda algorithm update) is a placement and audience setting, not a standalone strategy. Ask the agency when they use Advantage+ Shopping Campaigns, when they hold off, and whether they run Advantage+ Audience on retargeting. A good answer has account-specific conditions, such as whether prospecting carries 50 or more weekly conversions for at least 4 weeks, whether the creative library holds 8 or more distinct angles in rotation, whether the account is past 30 days of stable conversion signal, and whether retargeting audiences are large enough to keep their identity (typically a 180-day list above 100,000 users). A rehearsed pitch answer treats Advantage+ as always-on because Meta recommends it.

Advantage+ Audience on retargeting deserves its own callout. The setting blends cold and warm audiences inside the learning signal and degrades the retargeting pool over time. A diagnostic agency can walk through a specific account where they turned the setting off on retargeting and watched CPA improve. A pitch agency cannot.

When Advantage+ usually works:

• Advantage+ Shopping Campaigns often outperform manual structures for mid-funnel, but only when the creative set is diverse (8 or more distinct angles in rotation) and the account has a stable conversion signal (50 or more conversions per week per ad set, sustained over 4 or more weeks). On smaller accounts under $25K a month or in cold-start situations under 30 days of stable signal, manual structures still win.
• Advantage+ Audience on retargeting is a frequent misuse, because it blends cold and warm audiences in the learning signal and a diagnostic agency can name an account where turning it off improved CPA.
• Agencies overuse Advantage+ when it is the lower-effort default. Less account-specific work means a healthier agency P&L, but the client usually pays for that simplicity in CPA on prospecting and in degraded retargeting pools. Treat heavy reliance on Advantage+ across every account as a signal to ask harder questions.
• A bad answer: “We always run Advantage+, it is the future of the platform,” or “Meta recommends it.” Treating a placement setting as the entire strategy is a tell.

Ownership is the largest switching cost in a bad Meta engagement. The right setup keeps the Business Manager, pixel, Conversions API (CAPI) server container (the server-side connection that sends conversion events directly to Meta to recover signal lost when iOS blocks pixel firing), custom audiences, and ad accounts all on the client side, with the agency granted admin access. The wrong setup is any configuration where the agency owns one of those assets and reports out of its environment.

A good answer: “Everything lives in your Business Manager. We have admin access. If you fire us, you remove our access and nothing breaks.” A rehearsed pitch answer hedges with “we manage that for you,” “we handle the setup,” or “we would transfer everything at the end of the engagement.” Custom audiences and remarketing lists are non-trivial assets. A 180-day Meta remarketing list at scale is a material asset, and losing it on offboarding is a real cost. CAPI server container ownership matters for the same reason. If the agency built the server on their infrastructure, you pay their hosting forever or you lose the signal.

This is the last question, and it is the most revealing. By the end of an evaluation, the agency should have looked at your Meta account enough to name at least one specific change and the reasoning behind it. If they cannot, either the audit was shallow or the team is not right for the engagement.

A good answer names a specific change, the mechanism, and the expected metric movement. “Your prospecting campaigns have Advantage+ Audience enabled, which is blending cold and warm audiences in the learning signal. We would turn that off in week one. Based on similar accounts, that usually reduces prospecting CPA by 10 to 15% within 14 days.” A rehearsed pitch answer falls back on “we would start with a deeper audit to understand the account.” If the answer is good, it is also a free data point. You can act on it whether you hire the agency or not.

> Where most accounts leak first. For a 30-day audit map of the common first-30-day misses, see a 30-day audit map.

Use the ten questions across two meetings with the free audit in between. Meeting one is the filter. Ask questions 1, 2, 4, 6, and 9 to force specifics on creative motion, hit rate, measurement, retention, and ownership. The audit is delivered between the two meetings. Meeting two is the confirmation. Questions 3, 5, 7, 8, and 10 use the audit itself as the diagnostic and force the agency to commit to a specific 30-day account view.

Quick takeaways:

• Meeting one is a filter. Any agency that cannot articulate creative testing, hit rate, ROAS reconciliation, retention, or ownership is out, and there is no point in moving them through to the audit and wasting your time.
• The free audit happens between the two meetings. Grade it on specificity and dollar impact per finding.
• Meeting two is a confirmation. By the end, the agency has either shown diagnostic thinking or has not.
• A shortcut: if the audit is weak, the second meeting rarely recovers. Do not schedule meeting two for a clearly thin audit.
• A specific 30-day view means one concrete account change, the mechanism behind it, the expected metric movement, and a timeframe. “We would optimize the account” is simply not specific enough.

When the evaluation is done, book an in-depth audit with the agency that gave the most diagnostic answers. The audit itself becomes the final test. Three or more weak diagnostic answers across the ten questions usually means there is reclaimable spend in the account. Across 200+ audited accounts, 15 to 25% of non-brand spend has typically been recoverable in the first 90 days.

What questions should I ask a Meta ads agency before hiring them?

Ask about the creative testing framework on a live account, hit rate over 30 and 90 days, when they use creator versus advertorial versus brand-page creative, how they reconcile platform ROAS with actual revenue, who is actually managing your account day to day, their 12-month retention rate, what their free audit will actually find in dollars, how they handle Advantage+ on retargeting versus prospecting, who owns the Business Manager and the pixel, and what they would change in your account in the first 30 days. Each has a diagnostic answer and a rehearsed pitch answer.

How do I know if a Meta ads agency is actually good?

Three signals carry more weight than any pitch. First, creative testing hit rate on a live account: 5 to 7% winners over 30 to 90 days is normal at scale, and below 2% means the testing motion is not working. Second, reconciliation of platform ROAS to blended ROAS in the standard weekly or monthly report, with the gap named explicitly. Third, a free audit that names specific findings in dollars and tells you which findings you can fix without hiring the agency. If all three are strong, that is a strong signal the agency is worth a deeper conversation. If any one is weak, look at the next candidate.

What is a good client retention rate for a Meta ads agency?

Twelve-month retention above 90% is exceptional. Between 75% and 90% over the same window is acceptable. Below 75% over 12 months is a warning, and below 50% means the service model has a structural problem. Always anchor the question to a 12-month window so older accounts cannot dilute the read. Ask for the number directly. An agency that will not share it, or gives generic answers like “most clients stay multiple years,” is signaling something about its own churn. Retention is a trailing indicator of performance and fit combined.

Should I get a free Meta ads audit before hiring an agency?

Treat the free audit as the primary evaluation tool. A good audit names specific findings with a dollar impact on each line, identifies problems you did not already know about, and tells you which findings you can fix without hiring the agency. A rehearsed pitch audit uses generic observations and recommends a retainer as the only solution. The audit reveals more about agency quality than any case study, because the audit is the agency’s diagnostic thinking applied to your specific account, in front of you.

What contract terms should I avoid with a Meta ads agency?

Avoid multi-year minimums without performance outs, any contract without a 30-day cancellation clause, any language that gives the agency ownership of your Business Manager, pixel, CAPI server, or custom audiences, and any handoff language that does not guarantee data portability at offboarding. Month-to-month with 30-day notice is the cleanest form. Agencies confident in their work do not need long lock-ins.

What is the difference between platform ROAS and blended ROAS on Meta?

Platform ROAS is what Meta Ads Manager reports, calculated using Meta’s own attribution window, which includes view-through credit and overstates Meta’s contribution because of iOS gaps (conversions lost when users opt out of tracking via Apple’s App Tracking Transparency, introduced with iOS 14.5) and self-reporting. Blended ROAS is total revenue divided by total ad spend across all channels. A good agency volunteers the distinction and reconciles monthly. If your agency reports only platform ROAS, your actual performance is likely 50% or more lower than the dashboard shows.

What is the difference between a paid social agency and a Meta ads agency?

A Meta ads agency manages Facebook and Instagram ads only. A paid social agency usually covers Meta plus TikTok, LinkedIn, Pinterest, and sometimes Reddit or Snapchat. Platform expertise does not transfer automatically. A strong Meta team is not automatically a strong TikTok team. If most of your spend is on Meta, the Meta-specific depth matters more than channel breadth. Ask for case studies on the platforms where you actually spend. The strongest paid social agencies carry deep expertise within each channel they manage (Meta, TikTok, LinkedIn, and so on) rather than treating them as one motion.

How much does a Meta ads agency charge?

Monthly retainers for dedicated Meta management typically run mid-five-figure to low-six-figure depending on scope, channel mix, and creative production load. Percentage-of-spend pricing is common but creates an incentive to grow the budget rather than the business. Flat-fee or tiered-retainer pricing is generally cleaner and aligns agency incentives with performance.

Should I hire a big Meta ads agency or a small specialist?

Size matters less than seniority and account load. The better questions are who is actually managing your Business Manager day to day, and how many other accounts they manage. A small specialist with senior paid media leads at an 8:1 account ratio usually outperforms a large agency where junior strategists run execution at 15 to 20 accounts each. Size is a proxy for seniority at best, and often not even that.

How can I tell if my Meta ads agency is running on autopilot?

Autopilot here means an agency running off a fixed playbook without adapting to your account, not Meta’s automated tools. Check three signals. First, creative volume: an active Meta account should see 20 to 80 new ad variants per month depending on spend level. Second, the ad account change log: structural changes, audience updates, and creative launches should be weekly at minimum. Third, reporting substance: if the weekly report uses the same template every week with no new findings, new tests, or proposed changes, the account is on autopilot. A format can stay consistent across weeks as long as the substance inside it surfaces new diagnostics. Ask to see a recent week of activity in the Business Manager change log. Account names and client identifiers can be redacted.

What is a hit rate for Meta creative testing?

Hit rate is the percentage of tested ads that reach a winning threshold, where a winner is an ad that hits the account’s target CPA or beats the rolling 90-day account CPA by 20% or more, sustained over at least 7 days at meaningful spend. The benchmark is 5 to 7% of tested ads winning over a 30 to 90 day window at scale (based on 250 to 612 ads tested across 200+ audited accounts). A hit rate below 2% means the testing motion is not working: either the creative diversity is too narrow (hook variations on the same angle do not count) or the measurement setup is flagging false negatives. A hit rate above 10% usually means the definition of winner is too loose. Ask the agency to show the number from a live account.

What are red flags in a Meta ads agency pitch?

Watch for any of the following: refusal to share a hit rate on a live account, no numeric benchmark for creative testing, “data-driven” without a single number attached, reporting only platform ROAS without reconciliation to actual revenue, unclear Business Manager or pixel or CAPI ownership terms, free-audit findings without dollar impact per line item, long-term contracts with no performance outs, and a generic answer to “what would you change in my account in the first 30 days.” Any one is a warning. Two or more means pick another agency.

1. Meta for Business Help Center, “About Advantage+ Shopping campaigns”: https://www.facebook.com/business/help/advantage-plus-shopping-campaigns
2. Meta for Business Help Center, “About the Conversions API”: https://www.facebook.com/business/help/conversions-api