Meta & Facebook Ads
4 min read
Most launches fail structurally before a single creative is judged. Budget gets split across 6 ad sets, none of them reaches enough conversions to exit learning, and 3 weeks later you have 6 inconclusive results and an empty budget.
The free Claude skill campaign-architect sizes the structure so learning can actually finish, sets creative depth per ad set, and stages everything paused for your review.
This post covers the budget maths behind the structure, why consolidation beats fragmentation, and the naming convention that makes every other skill work.
Key takeaways
Each ad set needs a realistic path to roughly 50 conversions in a week, and the skill shows the budget maths before building anything.
When the budget cannot fund that across your planned ad sets, it cuts ad sets rather than starving each one.
3 to 5 creatives per ad set across at least 2 angles. Never a single creative gamble.
Everything is created paused, with a review date set, and the skill will not activate a campaign for you.
The budget check that decides the structure
daily budget per ad set ≥ (50 x expected CPA) ÷ 7
That formula comes straight from how the learning phase works: an ad set needs roughly 50 optimization events in a 7 day window before delivery stabilizes. If it cannot realistically get there, it never leaves learning, and everything it reports is unstable data you should not be making decisions on.
At a 40 dollar expected CPA that is roughly 286 dollars a day per ad set. If your total budget is 300 a day, you get 1 ad set, not 4. The skill runs this calculation first and states it plainly rather than quietly building a structure that cannot work.
Consolidate, don't fragment
Fewer ad sets with real budget each will beat many starved ones, every time. Fragmentation feels like thorough testing but it is really just spreading your data too thin to read.
So when the maths does not fit, the rule is to cut ad sets rather than cut budget per ad set. That is an uncomfortable recommendation when you have 5 audiences you want to try, which is exactly why having it enforced by a checklist helps.
The skill also states explicitly whether budget sits at campaign level or ad set level, because every downstream skill needs to know which level to act on. The default is campaign level for a single ad set launch, and ad set level when testing distinct angles that each need guaranteed spend.
3 to 5 creatives per ad set, across 2 angles
The structure needs creative depth to fill it, and a single creative launch is the most common reason a good structure underperforms. Vibelets generates on-brand product ads from your store in minutes.
The naming convention is infrastructure
Everything gets named by goal, angle, format, and date at every level. This looks like housekeeping and it is actually what makes the rest of the pack function.
The product winner finder maps spend to products through campaign names. The fatigue scanner needs format and angle to write iteration briefs. Without a convention, the highest attribution rung available to you drops to a blended estimate, and every product ranking becomes directional at best.
Name it right at launch and you never have to reconstruct it later, which is not a job anyone does willingly at scale.
How to set it up in Claude
Download the skill file from the skill's Google Drive folder, free, no email gate, then add it to Claude by uploading it as a skill under Settings or attaching it to a project or chat. Anthropic's Help Center guide to skills covers the upload path.
Bring 4 things: the product and goal, your daily budget, your target CPA or breakeven ROAS, and the creatives you plan to launch. The skill proposes the structure as a table and waits for approval before anything gets built, then sets the review date so nothing gets judged prematurely.
Conclusion
A starved ad set never learns enough to win. Size the structure to the learning phase, consolidate instead of fragmenting, launch with real creative depth, and stage it paused so you can look before it spends.
Build the structure so learning can finish
Free Claude skill, yours in 1 click.
Frequently asked questions
How many ad sets should a Meta campaign have?
As many as your budget can fund to roughly 50 conversions a week each, and no more. Divide 50 times your expected CPA by 7 to get the daily budget each ad set needs, then cut ad sets until the maths works.
How many creatives should I launch per ad set?
3 to 5, spanning at least 2 different angles. A single creative launch gives the system nothing to optimize between and gives you nothing to learn from.
Should budget sit at campaign level or ad set level?
Campaign level is the default for a single ad set launch. Ad set level makes sense when testing distinct angles that each need guaranteed spend. Either way, state which one you are using, because every downstream decision depends on it.
Why does the skill create everything paused?
So you can review the structure, budgets, and creative assignments before any money moves. The skill will not activate a campaign, and it will not set a budget you did not state.
Why does the naming convention matter so much?
Because later analysis reads it. Mapping ad spend to individual products, writing fatigue iteration briefs, and ranking products by profit all depend on being able to parse goal, angle, and format from the name.
Sources
Make ads that sell
your Products.
Generate on-brand product ads in minutes, no designer or brief required.
Trusted by 400+ growing brands