The structure below applies to any platform. The specific settings called out (ad set vs. campaign budget, objectives, the learning phase) are Meta’s.
The simple structure
Testing campaign
A lower-budget lane for new concepts, hooks, ICPs, formats, and offers. Its job is learning.
Winners campaign
A higher-budget lane for proven creatives and close iterations. Its job is scaling.
Where the budget lives: ad set vs. campaign
The setting that separates your two campaigns is where the budget sits. Meta renamed both options, so these are the labels you’ll see today:Set the basics (Meta)
For a Meta test, get these right and then leave them alone:- Objective: optimize for your real money event, usually purchase or subscription. If the account is new and lacks volume, Meta will ask you to optimize an earlier event (add-to-cart, lead) until it has enough data.
- Targeting: go broad. Set your market (geography) and let the creative find the audience instead of stacking interests by hand.
- Attribution: leave the default (7-day click, 1-day engage-through, 1-day view). More on why, and when to care.
What to test
Use a matrix instead of a pile of ads.
Structure each test as one concept per ad set, with 3–5 variations inside it. A concept is the idea or angle; a variation is one execution of it. The concept is your bet; the variations give Meta room to find the execution that lands. (Where concepts come from →)
Do not test ten unrelated ideas and call it a strategy. Pick a learning objective first.
Budget the test, and protect winners
- Give each concept ~2–3× your target CPA per day. That buys enough conversions to judge it in days rather than weeks. (Meta treats an ad set as out of its learning phase after about 50 events in 7 days. Smaller budgets run in “limited learning,” which is fine; just read thin numbers cautiously.)
- Never edit a winner. A significant change, like a big budget jump or new targeting, placements, objective, or creative, throws the ad set back into the learning phase and can wreck performance. If you have to change something, duplicate it and edit the copy.
Campaign setup workflow
Define the hypothesis
Example: “Parents convert better when the hook names exhaustion directly.”
Generate controlled variants
Ask Superscale for a batch that changes only the variable you want to test.
Launch with decision rules
Define what makes a winner or loser before the test starts: CPA, CTR, ROAS, hook rate, spend threshold, or another KPI.
Move winners forward
Scale winners, pause losers, and ask Superscale for the next generation of variants around the learning.
Reuse matrix
Before launch, decide which layers are new and which should be reused:Reusing an ad set or ad group also reuses its parent campaign because it belongs to exactly one campaign.
If you have no past data
Start with competitor research and customer language. Build a small testing matrix around likely pains and value propositions. Once results come in, connect that performance back into your context so the next batch is less speculative.Creative strategy
The testing matrix and concept engine behind the campaign structure.
Media buying on Meta
Budgets, the learning phase, scaling, and reading results.