Meta ads management

We run Facebook and Instagram accounts for brands spending $10,000 to $150,000 a month. The work is buying, creative testing, and scaling; the discipline underneath it is counting. Every account we manage reports two numbers each week: what Meta claims, and what your revenue shows.

The platform changed more in 2025 than in the five years before it. Management that still revolves around audience hacking is managing a machine that no longer exists.

Why did the old account structure die?

In December 2024 Meta announced Andromeda, a rebuilt ads-retrieval engine, and completed its global rollout by October 2025. Meta's engineering team reported a 6% recall improvement and 8% higher ads quality on tested segments. The practical consequence: the auction now does the audience selection, and it does it from your creative.

Meta then simplified campaign creation itself. In February 2025, Advantage+ Shopping Campaigns were folded into a single Advantage+ sales setup, and Advantage+ leads campaigns launched alongside. By mid-2026 those end-to-end Advantage+ tools passed a $75 billion annual revenue run-rate. Fighting that current with 30-campaign accounts stopped working; accounts structured for it saw roughly 8 to 10% performance gains in the panels that measured the shift.

How do we structure accounts now?

The structure we deploy is small, boring, and effective:

Consolidation is not a style preference. Each ad set needs enough conversion volume to exit learning, and a $30k budget split across twelve ad sets starves all twelve. The furniture retailer in our restructure write-up went from 14 campaigns to 2 and reached a stable CPA in 19 days.

Creative is the targeting now

Meta's own data science attributed about 56% of auction outcomes to creative quality, more than bid, audience, and placement combined. Its fatigue research found conversion likelihood drops roughly 45% after a person sees the same creative four times. Those two numbers define the job: volume and rotation beat the perfect ad.

We run a weekly testing calendar: four to six new concepts a month, judged on hook rate (3-second views over impressions, where 30% or better is a good score), hold rate, and cost per result. Winners get iterations; the rest get killed on schedule, not sentiment. Meta's generative tools now sit inside that cadence too. More than 8 million advertisers used them by early 2026, and they are useful for variations, though never for the concept itself.

What does scaling actually look like?

Budgets move against evidence. We hold spend until CPA is stable, then raise it roughly 20% per step so learning does not reset. Context matters here: ecommerce CPMs ran a $15.06 median over the last year, up 13% year over year, so flat spend quietly buys fewer impressions every quarter. A scaling plan that ignores CPM inflation is a shrinking plan.

The reporting stays on two ledgers throughout. Meta's attributed revenue sits next to your actual revenue, and when the gap widens we run the incrementality check before anyone celebrates. The method is documented on the process page.

What does it cost?

Management is a flat $2,600 a month up to $60k monthly spend, then 8% of spend. Typical agency pricing runs 10 to 20% of budget, which creates an incentive we would rather not have. The full fee schedule, including the $1,400 audit that starts every engagement, is on the pricing page.