A Facebook ads agency that wins on creative, not settings

What changed in Meta ads, and why it changes the agency you need

Five years ago a good Meta ads agency won with clever audiences and tidy campaign structures. That edge is gone. Meta's ad system, including the Andromeda retrieval model it rolled out across Advantage+, now decides who sees each ad based mostly on the ad itself. Broad targeting plus a varied set of creatives usually beats a dozen hand-built interest audiences.

In practice your creative is your targeting. A founder talking to camera reaches different people than a creator unboxing the product, and both reach different people than a static comparison chart. If every ad in your account is a variation of the same idea, Meta has fewer people to show it to and costs rise.

So we run Meta as a creative operation with a media buyer attached, not the other way round. Our creative strategy team plans distinct concepts, our short-form video team makes them, and the buyer reads the results every week.

Meta campaign types and formats we use

Names as they appear in Ads Manager in late 2026.

How our Facebook ads agency tests creative

Testing is the job. Here is the weekly loop.

1. 1

   Concepts before variants

   Each round adds new concepts (a new angle, format or person on camera), not just new colors on the same ad. Small tweaks rarely find new buyers under Andromeda.

2. 2

   Hooks on proven bodies

   Once a concept works, we cut 3 to 5 new opening seconds on the same body. Cheap to make, often a big lift.

3. 3

   Enough budget per test

   Meta's guidance is about 50 optimization events in 7 days for an ad set to exit learning. We size budgets to that, not to a round number.

4. 4

   Promote winners, retire losers

   Winners move into the scaling campaign. Ads that fatigue get refreshed with a new hook or a new creator.

5. 5

   Fix the page too

   If click-through is strong but sales are weak, the problem is after the click. We fix the landing page or funnel before buying more traffic.

What Meta ads cost in 2026

Ranges are third-party 2026 benchmarks across Facebook and Instagram. Your category and season move them a lot.

The test month figure is our planning math from the learning-phase target, not a Meta minimum. Smaller budgets work if you optimize for a cheaper event like add to cart or lead.

How a Meta ads agency should measure results

Meta's own reporting is a starting point, not the answer. Since iOS privacy changes, in-platform numbers are partly modeled, and Meta tends to claim credit for sales that other channels helped start. We set up the Pixel and Conversions API properly on day one, then check three numbers each week: Meta-reported cost per purchase, blended revenue against total ad spend from your store, and what buyers say in a one-question post-purchase survey.

When the numbers disagree, we trust the store. For bigger budgets we also suggest a holdout test, pausing Meta in one region for a few weeks, to see how much revenue it really drives. You get one weekly report with all three numbers side by side, and a plain note on what we are changing next. If you want the full sequence, our campaign launch guide walks through it.

Meta ads mistakes we fix most often

• Ten ad sets with tiny budgets, none of them leaving the learning phase.
• Twenty near-identical ads that Meta treats as one idea.
• Interest stacks and lookalikes fighting Advantage+ instead of feeding it creative.
• Broken Conversions API or duplicate events, so Meta optimizes toward bad data.
• Editing budgets by large jumps every day, which keeps resetting learning.
• Judging success on in-platform ROAS alone, with no check against store revenue.
• Sending paid traffic to a homepage instead of a page that matches the ad.

What Lumiflo costs

Ad spend is paid by you, straight to the platforms. Creator and influencer fees are billed at cost.

See what each plan includes.