How many ad sets and ads should a new campaign have?

Structure is where small budgets go to die. Not because the wrong structure performs badly, but because a fragmented structure cannot produce an answer — you spend the money and end up knowing nothing.

The rule underneath everything below: every ad set needs its own 50 optimisation events per week. Structure is the arithmetic of dividing your budget so that remains true.

Start from the budget, not from the idea

The number of ad sets you can support is not a matter of taste. It is division.

max ad sets = weekly budget / (50 × cost per optimisation event)

At $300 a week and a $4 add-to-cart cost, one ad set needs $200 a week to learn. You can support one, and you are kidding yourself if you run three.

At $300 a week optimising for purchases at $30 CPA, one ad set needs $1,500 a week. You can support zero — which is the forecast telling you to optimise for a cheaper event, not to give up. The budget calculation goes through that trade in detail.

This is the whole reason “concentration beats spread” is true at small budgets. Two ad sets at $15/day both stay in learning forever and neither produces a usable read. One at $30/day gets you an answer.

The starting structures that work

Under $1,000/month.

• 1 campaign, conversion objective
• 1 ad set, broad targeting, optimising for the cheapest event you can reach 50 of
• 3–4 ads, genuinely different creative angles

$1,000–$3,000/month.

• 1 campaign, conversion objective
• 2 ad sets: broad prospecting, plus one lookalike if you have a seed worth using
• 3–4 ads each
• Optionally a small retargeting campaign once you have enough traffic to fill a pool

$3,000+/month.

• 1 prospecting campaign with 2–4 ad sets, split by audience type rather than by micro-interest
• 1 retargeting campaign, separate budget
• 4–6 ads per ad set, refreshed on a schedule rather than in a panic

Notice that the ad set count grows slowly and the creative count grows faster. That is the correct ratio, because creative is where the returns are.

Splitting by audience type, not by micro-interest

When you do add a second or third ad set, split on something structural:

• Broad versus interest-based
• Prospecting versus retargeting
• One country versus another, where language or shipping differs
• Lookalike from purchasers versus lookalike from site visitors

Do not split into “yoga”, “pilates” and “fitness” as separate ad sets. Those audiences overlap heavily, they compete with each other in the auction, and you have tripled the events you need with no new information in return. If you want those three interests, put them in one ad set with OR logic and let delivery sort it out. Targeting from zero covers why narrow stacking costs more than it returns.

How many ads per ad set

Three to five. Below three you are not testing anything; above six the budget per ad gets too thin for delivery to distinguish between them, and Meta will concentrate spend on one or two anyway.

The important part is that the ads are different, not variations. Four crops of the same photograph with the same headline is one ad tested four times. Four angles — product in use, problem framing, social proof, specific benefit — is a real test.

When one ad clearly wins, do not immediately delete the others. The loser gives you the contrast that tells you why the winner won, and that is what transfers to the next campaign.

Campaign budget or ad set budget?

Use ad set budgets while you are learning. You want control over exactly what each audience gets, and you want to be able to read each one independently.

Move to campaign budget optimisation once you have two or more ad sets you are confident in and you want Meta to allocate between them dynamically. Doing it earlier hands the allocation to an algorithm that will starve one ad set before it ever produced data — which is a fine outcome for spend efficiency and a terrible one for learning.

Do not restructure to fix a creative problem

The most common expensive mistake: results are poor, so the campaign gets rebuilt with new ad sets and a new split.

Restructuring resets the learning phase on everything and destroys the comparability of every number you have collected. If CTR is weak, the creative is the problem and no arrangement of ad sets fixes it. Change one thing, wait three days, read.

---

AdPilot’s campaign generator sizes the structure from your actual budget: how many ad sets that budget can keep above the learning threshold, how to split them, and how many ad variants to write for each.