The short answer
The structure that fits Meta in 2026 is small: one consolidated scaling campaign for proven ads and one testing lane for new concepts. The data backs the concentration: among 125 brands with 5+ winners on our board, the top ad carries a median 32% of the brand's estimated spend and the top three carry 68%.
The structure in one table
For most direct-to-consumer brands spending from a few thousand to a few hundred thousand dollars a month on Meta, this is enough:
| Campaign | Purpose | Targeting | What goes in it |
|---|---|---|---|
| Scaling | Spend most of the budget on what works | Broad, Advantage+ audience | Proven winners and their variants, 1 to 3 ad sets |
| Testing | Give new concepts a fair read | Broad, same hard limits | New concepts, a fixed share of budget |
| Optional: separate market or product | Keep different economics apart | Per market | Only when margin, country or catalog really differs |
That is it. No interest stacks, no ten lookalike ad sets, no campaign per creator. "Simple beats clever" is not a slogan here. It follows from how Meta's delivery learns.
Why consolidation wins
Three facts drive it.
- Learning is per ad set. Meta's help center says an ad set generally needs about 50 optimization events after its last significant edit to stabilize. Splitting a budget across ten ad sets means ten separate counters, each slower to fill.
- Targeting is mostly automatic. With Advantage+ audience, the system can expand beyond your selections. Separate ad sets for separate interests often end up reaching the same people.
- Creative picks the audience. Since Andromeda, the system matches people to ads based on the creative. Segmentation moved from the ad set to the ad.
So the structure should give the system as many conversions per ad set as possible and as many distinct concepts per ad set as you can produce.
What winners' portfolios look like
We can't see campaign structures in the Meta Ad Library. We can see how brands spread estimated spend across the winners they keep live, which is the outcome a structure produces. These numbers cover the 125 brands with five or more winners in our October 9, 2026 snapshot.
32%
of a brand's estimated spend sits on its single top ad (median)
68%
sits on its top three ads
189 days
median gap between a brand's oldest and newest live winner
56 of 125
brands run a 180+ day winner next to one launched in the last 90 days
Spend is concentrated. Even brands with five, ten or twenty winners put about two thirds of their estimated spend behind three ads. That is what a scaling campaign does on its own: it finds the best ads and feeds them.
And portfolios span ages. 83 of the 125 brands still run a winner that is at least six months old; 96 have one launched in the last 90 days. A healthy account looks like a few long-running pillars plus a steady flow of new ads, which is why it needs both a scaling home and a testing lane.
There is also a pattern in intensity. Estimated spend per day, averaged over each ad's life so far, falls as ads age:
| Days live | Ads | Median estimated spend per day |
|---|---|---|
| 30 to 59 | 172 | ~$3,100 |
| 60 to 89 | 529 | ~$2,550 |
| 90 to 179 | 1,037 | ~$1,620 |
| 180 to 364 | 367 | ~$1,140 |
| 365 or more | 338 | ~$560 |
New winners are pushed hard; old ones keep running at a lower, steady rate. Don't pause an old winner just because a new one took most of the budget. It may still be finding buyers the new one doesn't reach. See creative fatigue for how to tell.
Pillars and newcomers from one brand
Jones Road Beauty shows the pattern in miniature: a long-running video pillar and a newer static being scaled with many versions.
I don't want to wear foundation every day, so I do this instead.
- Est. spend
- $3.8M
- Days live
- 608
- Format
- Video, 59s
The pillar. Twenty months live on one version, with a second cut of the same opener also running.
The easiest way to find your shade
- Est. spend
- $428K
- Days live
- 78
- Format
- Image
- Variants
- 12
The newcomer. Twelve versions in 78 days, sending people to a shade quiz. This is a concept graduating into scale.
Misfits Market is the extreme case of a pillar: one video carrying 84% of the estimated spend across the brand's five live winners, for more than a year and a half.
It got stuck in the branch that it was growing on, and then it also has a line from where it was strung on the plant, but still perfectly edible.
- Est. spend
- $22M
- Days live
- 581
- Format
- Video, 79s
- Variants
- 8
A pillar with eight versions. A scaling campaign built around one ad like this needs a testing lane next to it, or nothing else gets spend.
Where new concepts go
Two options work. Pick one and stick to it.
Option A: separate testing campaign. New concepts launch in their own campaign with a fixed budget, sized so each concept can spend a few times your target CPA within a week or two. Each concept gets a fair read without competing against a pillar. Winners move to scaling as new ads. This is the safer default.
Option B: new ad set inside the scaling campaign. Each batch of concepts gets its own ad set in the scaling campaign. Simpler, but the campaign budget will drift toward the proven ad set, and new concepts can starve. Works best at high spend, where even a small share is a real test.
Either way:
- Launch in batches. Four to six distinct concepts at a time is enough for most budgets. See how many ads per ad set.
- Judge on spend, not days. A concept needs to spend a few times your target CPA before you call it. Our creative testing budget guide shows the math.
- Graduate as a new ad. Copy the winning ad into the scaling campaign. Avoid significant edits to the scaling ad sets themselves, since those can restart learning (learning phase).
- Keep the pillars. Don't pause a long-running winner to make room. Let the budget decide.
What to leave out
- Interest-segmented ad sets. They overlap under Advantage+ audience and split learning.
- One campaign per format or creator. Format is a creative choice; put formats side by side so delivery can choose.
- Frequent duplication to "reset" an ad. It restarts learning and usually reaches the same people.
- Retargeting campaigns by default. Broad campaigns already reach warm audiences. Add one only if you have a distinct offer for them.
If you want to see how competitors balance pillars and newcomers, follow them on Ad Radar and compare days live across their winners. The pattern above (a few old ads carrying the account, a stream of new ones behind them) shows up brand after brand.
Figures marked as estimated spend come from Ad Radar's model of engagement on public Meta Ad Library ads. They are estimates, labeled as such, and are best used to rank ads against each other.