By December 31, 2027, pull 20 proposals for US Meta accounts spending $10k to $50k a month: I expect at least 15 to lead with a flat retainer or a flat base plus a CPA kicker. Straight percent-of-spend pricing will be at the bottom of the page or gone.

I say that as someone who billed that way for years. I used to tell clients the percentage kept incentives aligned: when you grow, we grow. I was wrong. The work stopped scaling with spend a while ago, and Meta’s automation absorbed much of the audience, placement and bid work we used to charge for. What is left is creative volume, offer strength and signal quality. None of those automatically costs twice as much when you double the budget.

The fee stopped matching the work

Say you spend $20k a month and pay 15%. That is $3,000 for management. Double your spend to $40k and the fee becomes $6,000. Did the work double? The budget field changed. Meta’s system still selects audiences and placements and paces delivery. I used to earn that increase by rebuilding interest stacks and placement breakdowns by hand. That job barely exists now.

The shift is mechanical. Advantage+ Audience is now the default for Sales, Leads and App objectives, with interest and age inputs treated as suggestions rather than hard limits. Detailed-targeting exclusions were removed on March 31, 2025, with unmigrated ad sets stopping delivery on January 15, 2026. Meta’s stated end state goes further: supply a product image or URL and a budget, and its system generates creative, selects audiences and placements, and optimizes delivery. The same account of that plan reports that the GEM foundation model drove a 5% lift in Instagram conversions and 3% in Facebook Feed conversions.

That does not mean there is nothing left to manage. Automation takes on campaign setup, audience selection, placement choice, creative variation testing, budget allocation and real-time bids. The buyer still owns signal quality, creative direction and guardrails, measurement design and cross-platform interpretation. The job shifts from pulling every lever to checking what the machine learned and whether it learned the right thing.

That is where the money should move. A survey of more than 200 DTC agencies describes the remaining work as strategy and structure, creative testing and iteration, reporting, and proactive communication. It singles out creative as worth top-of-market pay because one breakout ad can swing $20K to $50K in monthly profit. I have seen that math hold. A $20k account does not need a prettier interest stack. It needs more swings at a winner.

Read the monthly report accordingly: if it leads with audiences tweaked and bids managed, ask what happened beyond Meta’s own delivery work. If it leads with ads shipped, winners scaled and signals fixed, you are looking at work worth discussing.

Prediction 1: Flat fees lead Meta proposals under $100k by Q4 2027

Meta-only management without creative now benchmarks at 3% to 10% of spend, falling as spend rises, or $2,500 to $5,000 a month for accounts spending under $100k. Add creative, CRO or Google, and the range rises to 15% to 20%. The distinction matters: I am not claiming a creative team costs the same as a manager overseeing automated delivery.

But a bare percentage still fails the smell test. At 10%, a $15k account pays $1,500; an $80k account pays $8,000. If the scope is substantially the same, the larger budget does not explain the larger fee. More than 200 agencies surveyed cluster around a flat retainer or a retainer plus a percentage. The survey describes percent-of-spend as fair only with a cap and a CPA, ROAS or MER target attached.

The replacement is already on the shelf: groas prices Meta Ads at $1,499 a month, regardless of ad spend, with no cut of spend. That is a better fit for an account paying an ever-larger management fee for delivery work Meta already handles.

What proves me wrong: collect 20 US Meta agency proposals in December 2027 for accounts spending $10k to $50k a month. I expect at least 15 to lead with a flat fee or a flat base plus a CPA kicker. If 10 or more instead lead with uncapped percent-of-spend and no flat option, I missed.

Prediction 2: Creative output replaces audience counts by mid-2027

When distribution becomes easier to automate, agencies need more good material to distribute. That makes creative output a more useful proposal number than the number of audiences someone promises to manage.

A working benchmark is about one new ad per $3,000 of monthly spend: roughly 10 ads at $30k and 33 at $100k. In an analysis of 578,750 creatives across 6,015 accounts, about 5% became winners. At that rate, 40 ads a month yield about two winners; five yield close to none. More volume is not a substitute for a strong offer or direction. It does give the system more chances to find what works.

By mid-2027, I expect serious proposals to quote ads shipped per month the way they used to quote audiences managed. The first page might say 12 new ads per month, three iterations per winner, losers killed in 72 hours. Those are commitments a buyer can check, not a slide about audience architecture. groas already frames its Meta Ads offer around new ads tested and winners scaled. An agency selling four ads a month against $50k in spend is selling lottery tickets one at a time.

What proves me wrong: if December 2027 proposals still lead with audience matrices and bury creative output on page four, I missed. Ask how many new ads the agency will ship before you admire its targeting diagram.

Prediction 3: CPA pricing grows through 2027, then runs into attribution

I get the appeal. You are tired of an agency earning more whenever you spend more, so paying per acquisition feels cleaner. Expect more of those offers through 2027. CPA deals are built around a target CPA and margin; revenue share commonly runs 3% to 10% of tracked sales. But pure-performance deals carry higher effective rates to cover losing accounts and tend to favor proven offers. An agency taking that risk will be selective. You are not necessarily buying better alignment. You may be buying a handicapper who only bets when it likes the cards.

Then you have to agree on what counts. On January 12, 2026, Meta removed 7-day view and 28-day view from the Ads Insights API; reported conversions could drop 15% to 40% overnight without a campaign change. For B2B sales cycles of two to four weeks, the same account describes 20% to 40% fewer attributed conversions and identifies 7-day click plus 1-day view as the new standard. It also says pixel-only setups can miss 30% to 60% of conversions, making Pixel plus CAPI the baseline.

A CPA contract turns those measurement choices into money. If the agency defines conversions itself or controls a pixel and CAPI setup you cannot inspect, you can end up paying against numbers you do not control.

What proves me wrong: if CPA contracts in December 2027 commonly specify one shared pixel, one written attribution window and client-owned CAPI, without the measurement fight I expect, I overstated it. Until then, do not sign outcome pricing without owning the measurement.

Prediction 4: Audience strategy loses its proposal page by December 2027

I used to sell audience strategy as a line item: interest stacks, lookalike ladders, exclusion architecture, Feed-versus-Reels placement splits. Clients paid for the diagram, and I drew it carefully. When audience inputs become suggestions and the system decides who sees what, that diagram starts to look like a map of a city that rebuilt its roads overnight.

There is still work in setting guardrails and checking the signals that guide delivery. There is less reason to bill for an elaborate manual audience plan as if the platform will follow every line. That is why I expect the proposal section called audience strategy to shrink to a sentence, while creative and measurement get the space it used to occupy.

What proves me wrong: if December 2027 proposals still open with a three-page audience matrix and win pitches on its strength, I missed. If your proposal bills audience research as week-one work, ask what the agency will research that Meta is not already testing live. Pay to keep the plumbing clean, not to admire its blueprint.

Prediction 5: Signal quality separates accounts by December 2027

Automated targeting depends on the events it receives. If those events are incomplete, the system makes decisions on incomplete information. That is why signal quality from Pixel plus CAPI is one of the biggest remaining levers.

I have watched accounts with identical budgets and ads diverge by 40% on CPA for one reason: one sent clean purchase events, while the other relied on browser-only guesses. Pixel-only setups can miss 30% to 60% of conversions. A weak signal gives the model a partial account of what happened. It can then buy the wrong impressions with impressive confidence.

This is not a pitch to stare at a tracking score while bad ads keep running. It is a reason to check the inputs before declaring the algorithm broken or raising the budget. Ask for event match quality and the deduplication rate before you ask why ROAS moved.

What proves me wrong: open Events Manager in December 2027. If pixel-only accounts without CAPI still match deduplicated setups on event match quality and CPA stability, I missed. If your agency cannot show you the signals it sends, start with the plumbing before you pay for another targeting plan.

Who should ignore these predictions

This will not fit everyone. If you spend $500k a month, have an in-house creative team shipping 100 ads a month and employ a media buyer who only runs Meta, a flat external fee will not replace that machine. You need custom testing infrastructure, not a simpler invoice. The same goes for a proven direct-response offer with a CPA partner that already beats your target and owns the risk. I used to push every account toward the same model. That was lazy. If spend and creative output already scale together inside your team, keep them together.

Under $100k, the arithmetic often runs the other way. Say you pay $4,500 on $30k in spend and receive six ads plus a targeting report. That is $750 in management fees per ad shipped before accounting for the agency’s other work. You should know what that other work is. A flat $1,499 offer that tests new ads and scales winners has a clearer answer than a fee that rises just because your budget does. It stops charging a percentage for delivery work the platform already performs.

What to ask your agency this quarter

If I am right, you do not need another agency deck. You need four answers on your next call. I ask these with every account I review; good operators can answer in numbers.

  1. How many net-new ads did you ship last month per $3,000 of my spend, and how many became winners? If output is under half the working benchmark, look at the creative pipeline before blaming delivery.
  2. What are my Event Match Quality and deduplicated purchase count in Events Manager? If CAPI is missing, fix the signal before you raise budget.
  3. Which written attribution window prices my CPA? If you cannot agree on the window and keep measurement client-owned, do not sign outcome pricing.
  4. What exactly am I paying a percentage for? If the answer is audiences, placements and bid tweaks, ask which parts Meta already handles.

Save one proposal from this quarter. In December 2027, pull 20 more and count how many lead flat and how many lead with audience matrices. I will check it the way I check ad tests: dated and counted. If flat wins and the matrix is gone, the percentage model did not evolve. It ended. And if you want the flat version now, groas runs Meta Ads at $1,499 a month, with new ads tested and winners scaled, without taking a cut of your spend.