Ad Set Split
Updated Saturday 17 October 2026, 7:58pm Pacific
Meta's dashboard is warning that our ad sets are "fragmented" and suggests combining them. This page is the Meta ads committee's answer on whether to keep our setup of one ad set per product, and it matters because merging would change how our money gets moved between products.
Meta ads: keep one ad set per product (BROQ, NMN-99, SPMD-99 and multi-product), even though Meta's dashboard suggests combining them. The committee backs this: 7 of the 11 members with a view keep the split by product, and not one would merge the pairs Meta names.
Meta's "31% lower cost" is a model's estimate, not something it measured on our account. The same $69.29 to $48.50 cost per purchase sits on all four of its suggestion cards, and Meta's own help says the score "does not reflect your actual or future performance."
Merging wouldn't get us out of Meta's learning phase anyway. Meta wants about 50 purchases a week in an ad set before it stops "learning", and all four ad sets together made 26 last week, about half that.
What we'd lose by merging is real. Today our automatic budget rule moves money between the products at 12:01am every other night, on new-customer sales only; in one merged ad set, Meta would move the money itself on its own purchase count, returning customers included. Every ad would also restart its learning, including the SPMD-99 "holy grail" video that carries that ad set.
Merging would also crowd a single ad set. With the seven new slideshow ads, one ad set would hold 34 ads and reach Meta's cap of 50 in about three weekly batches.
Within each ad set, there's nothing to do yet. The biggest is SPMD-99 at 9 ads (11 with the slideshows), well inside what most members run. Keep adding weekly batches of truly different ads, and look again when any ad set passes about 20.
Two things worth knowing. Five members usually give the whole campaign one budget instead of one per ad set, so that's the change to look at if our budget rule is ever retired. And nobody should press Meta's "Apply now" or "Set up auto-apply": Meta would keep the ad set it judges best, move the others' budgets into it and turn them off.
The choice
The three choices are in the brief's Questions. Nothing changes in Meta until you and Nick answer.
- Recommended: keep one ad set per product. Four ad sets, the automatic 12:01am budget rule keeps moving the money on new-customer sales, Meta's suggestion cards stay unapplied, and we look again when any ad set passes 20 ads.
- Keep the four ad sets, with one budget for the whole campaign. Meta then moves money between products itself, on its own purchase count (returning customers included).
- Merge everything into one ad set, as Meta suggests. Meta picks which product gets the money, every ad restarts its learning, and the ad set hits Meta's 50-ad cap by early November.
The full review
The question put to the committee (Mike, Saturday 3:43pm, word for word)
"Run this by the meta ads committee: Are we definitely doing the right thing to have the ad sets separated the way we do? I suppose if they weren't separated by product, they'd need to be separated in a more arbitrary way, since we're going to quickly get to (or might already be at) too many ads for a single ad set. So probably leave as is. It's just that meta is saying fragmentation, so i want them to weigh in on it."
What Meta says (checked today, about 3:50pm)
- The cards: "31% lower cost per purchase potential if you combine 2 similar ad sets affected by fragmentation." Four cards of two ad sets each, every one showing $69.29 against $48.50.
- It's a model, not a measurement. Meta: the score "does not reflect your actual or future performance." $48.50 is exactly $69.29 less 30% on all four cards. Our own 7-day figure is $82.19 per purchase ($2,136.86 for 26).
- Not applying is allowed. "You should judge performance based on your business goals." Dismissing a card leaves the score as it is.
- What "Apply" does: "We keep the best performing ad set... and expand it," taking the others' audiences and budgets and turning them off, and "we delete duplicate ad creative." Best is judged on Meta's own cost per result, returning customers included.
- Our ads don't bid against each other. Meta picks one per auction. The risk is an ad set that can't spend its full budget.
- Ads per ad set: hard cap of 50; Meta gives no recommended number.
Our account today
- One Sales campaign with a budget on each ad set: SPMD-99 $137.50, multi-product $93.60, BROQ $60.56, NMN-99 $45.00 ($336.66 a day).
- Why Meta sees "different audiences": only multi-product has two extra targeting settings. Three cards pair multi-product with a product; the fourth pairs NMN-99 with BROQ. Matching the settings would be a targeting edit that resets learning, for no gain.
- Last 7 days, Meta's count (all customers): SPMD-99 15 purchases, multi-product 4, BROQ 4, NMN-99 3. All four in learning; 26 together.
| New customers, Wed 14 to Fri 16 Oct | Without spillover | With |
|---|---|---|
| SPMD-99 | 1.00x | 1.26x |
| NMN-99 | 0.71x | 0.90x |
| Multi-product | 0.22x | 0.27x |
| BROQ | 0.20x | 0.25x |
| Overall | 0.57x | 0.72x |
Multiples of break-even. Break-evens without spillover: SPMD-99 1.71, BROQ 1.72, NMN-99 1.73, multi-product 1.59.
The members on the split
Nick Theriot: keep the split Different product categories get their own campaign; NMN, spermidine and sulforaphane are categories, not variants.
Taylor Holiday: keep the split Splits by offer; each product and the stack is its own offer.
Ben Heath: don't mix products in one ad set One campaign per product range. For a small account he'd sell fewer products before merging different ones.
Andrew Faris: keep products apart Wants purchases concentrated, but ad sets matched by order size and margin; mixed, Meta overspends on the cheaper product.
Nick Shackelford: keep ads spread out His commonest audit finding is "too many ads in too little of ad sets". One ad set of 34 is that.
Ash Melwani: close to where we are His supplement brand runs "two or three ad sets with like 10 to 15 ads".
Jared Robinson: don't restructure now Structure is "an afterthought" while the creative isn't right.
Charley Tichenor: consolidate further "Simple scales and complex fails." He'd put everything behind one hero offer (SPMD-99), which your rule of 12 October (one ad set per product, plus one multi-product) rules out.
Jon Loomer: leans to fewer ad sets Splits only for products that each have the budget to get about halfway to 50 a week. Only SPMD-99 is near that.
Sam Piliero: different structure One campaign budget and a new ad set for each batch of ads. The 12 October rule of one ad set per product rules that out.
Chase Chappell: mixed Ad set budgets suit small accounts, moving to one campaign budget past about $200 a day.
No opinion on this: Cody Plofker, Dara Denney, Sarah Levinger, Savannah Sanchez, Alex Cooper. Call-ins, quoted not seated: Barry Hott ("Consolidate or die."), Andrew Foxwell (one campaign gives clearer signal), Brad Ploch (one campaign per product).
Ads per ad set
Six members run 10 to 50 (Heath, Chappell, Robinson, Holiday, Faris, Melwani); Theriot says the count doesn't matter. Only Piliero (2 to 8) and Tichenor (cuts past 8) would trim at our sizes; Loomer and Shackelford prefer few. At about 2 new ads per ad set a week, SPMD-99 nears 20 around 20 November. Then, before each batch, ads that have run two weeks or more with almost no spend come off first (recorded for then, not decided). Every new ad must be truly different.
Where they agree
- Don't merge different products into one ad set.
- The fix for thin data is more sales per ad set, not new ad sets. The automatic 12:01am budget rule already moves money toward the one that sells.
- Ad count isn't a problem yet.
Where they disagree
- Campaign budget or ad set budgets. Five default to one campaign budget; Chappell keeps ad set budgets under about $200 a day. Ours are per ad set because the automatic budget rule moves each product's money on new-customer sales. One campaign budget is the second choice above.
- How few ad sets. Tichenor and Loomer would run fewer than four; the rest accept a split by product.
What would change the answer
- SPMD-99 passing about $200 a day (Chappell's point for one campaign budget).
- Multi-product falling short of its budget, which would be overlap. Checked after the review: it spends its whole budget ($91.75 of $93.60 on Friday, 98%).
- One product ad set nearing 50 sales a week.
- Meta's purchase count lining up with new customers; then one campaign budget is worth a fresh look.
- Any ad set passing about 20 active ads; back to the committee on which come off.
Recorded, not in the plan
- Tichenor's one hero offer, Heath's one-product start, Piliero's ad set per batch: against the 12 October rule of one ad set per product, plus one multi-product.
- Matching multi-product's targeting to the others: a learning reset for no gain.
- Meta's flexible ads to cut ad counts: no member backs it, and it would hide which video sells.
- "Apply now" and auto-apply: never.
Sources: the Meta ads committee's review of Saturday 17 October and the report on it from Claude's Meta ads session (3:55pm Pacific), from Meta's help pages, the Meta ads knowledge base and our budget log. Nothing in Meta was changed.