How many creatives should one ad set have?

Between four and six, genuinely different from each other (angle, format, voice). Fewer and there is nothing to compare; more and the budget spreads too thin to learn anything about each.

Is Advantage+ mandatory?

No. You can turn off or narrow most Advantage+ options. But you need a concrete reason — an exclusion, a real constraint — not just discomfort with letting the system decide.

Why did cost go up after I doubled the budget?

A large budget change sends the ad set back into the learning phase and forces it to find more people inside the same audience. Increase gradually, in moderate steps, every few days.

How a Meta Ads account is structured in 2026: campaign, ad

The situation

You open Ads Manager for the first time and see three levels, dozens of settings and a blue button urging you to let Meta decide everything. Every business owner asks the same first question: "Which interests do I pick?" It is the wrong question for 2026. Interests were the main lever in 2018. Today the lever is the creative, and the account structure has to be built around it.

The three levels

The campaign — you choose the objective: awareness, traffic, engagement, leads, app promotion or sales. The objective tells the system which event to buy. If you want orders, you choose sales and optimise for Purchase, not traffic — cheap clicks are not customers.
The ad set — who, where, with what budget and which event you optimise for. Audience, placements, schedule, budget (unless set at campaign level), the conversion event and the pixel or dataset.
The ad — the creative: video, image, carousel, primary text, headline, call to action, destination.

The structural rule that holds in 2026: few ad sets, many different creatives. Every separate ad set splits the budget and has to exit the learning phase on its own. Ten ad sets at 30 lei a day each never learn. One ad set at 300 lei a day with six creatives does.

Advantage+ vs manual

Meta has pushed automation under the Advantage+ name: automatically expanded audiences, automatic placements, a campaign-level budget distributed by performance, and for sales and leads, a campaign flow where you choose little and the system does the rest.

When to use it: you have a pixel or Conversions API with clean events, at least a few dozen conversions a week, and several creatives to feed it. The system finds the buyers on its own.

When to stay manual or narrow down: you want to exclude existing customers from an acquisition campaign, you have real geographic or age constraints (a clinic in a single city, an adults-only product), you are testing a message on a specific audience, or the budget is small and you want control over one segment.

Broad or interests?

In 2026, for most accounts with conversions, broad (country, age, maybe gender) plus good creative beats a stack of interests. The delivery system uses pixel signals and how people react to the ad far better than the label "interested in fitness". Interests remain useful as a guiding signal inside Advantage+ audiences and in new accounts with no data.

The creative does the targeting. A video that opens with "If you run an online store in Romania and ship by courier…" selects its own audience: people who do not recognise themselves scroll on, and the system learns from who stops. That is why creative diversity matters more than audience diversity: different angles (problem, proof, offer, comparison), different formats (vertical video, static image, carousel), different voices (founder, customer, demonstration).

Budget and the learning phase

Every ad set enters the learning phase when you launch it or change something significant (budget by a lot, audience, creative, event). The system explores. Meta indicates it needs roughly 50 optimisation events within a week to stabilise delivery. If it does not get there, you see "Learning limited" — delivery stays unstable and usually more expensive.

Practical consequences:

The daily budget per ad set has to allow for those events. If a Purchase costs 100 lei, an ad set at 50 lei a day will never learn. Pick a more frequent event (Add to cart, lead) or consolidate the budget.
Change nothing in the first days. Every major edit resets learning.
Adjust budgets gradually; do not double overnight.

What to look at in the first 7 days

Not ROAS. It is too early and the data is thin. Look at the upstream signals:

CPM — what it costs to reach 1,000 people. If it is far above your other campaigns, the audience is too narrow or the creative is being penalised.
Hook rate — 3-second views over impressions. It tells you whether the first seconds stop the thumb.
Link CTR — whether the people who stopped also click.
Cost per result on the upstream event (add to cart, lead), watching the trend, not the absolute value on day one.
Frequency — above 2–3 in a week on a small audience means you have exhausted it.
Which creatives receive budget — the system shifts budget towards what works on its own; the ones that get nothing in 3–4 days you replace, not force.