Meta ad creatives: how many, which formats, how to test them

On Meta today, the creative decides who sees an ad and who responds. Audiences are getting broader, the algorithms find the people themselves, and the ad sets the direction. That is why creative testing at ad.boutique is not an extra, but the core of every campaign. More about our production on the page content production.

In short

• 4 to 6 concept creatives per ad set.
• Three formats: 1:1, 4:5, 9:16.
• Concept beats variant: test different ideas, not the same idea in five colours.
• Monday review: evaluate every week, scale winners, replace losers.
• AI speeds up variants; final creatives are curated and labelled on request, and we always comply with statutory labelling obligations (for example Art. 50 of the AI Act).

Concept and variant

A concept is the idea behind a creative: which person, which problem, which proof, which tone. A variant is the same idea with a different hook, a different length or a different format.

Meta recognises very similar ads and partly treats them as one. To the system, five variants of the same idea are closer to one creative than to five. That is why we test concepts first and then keep developing the winners.

Psychological levers are a good starting point for different concepts: social proof for the UGC creative, the carpenter effect for the product in use, the anchoring effect for the order in the carousel. We explain all ten in 10 psychological effects in marketing.

Formats: 1:1, 4:5, 9:16

We produce every concept in three formats:

The most important rule: the most important thing belongs in the centre. Copy, logo and product must not be covered by buttons or profile pictures. And videos have to work without sound, so with subtitles.

Six creative types

These six types cover most campaigns. We mix several in every batch.

1. Talking head: a person speaks directly into the camera, a founder, an expert or a customer. Strong for trust and explanation.
2. Case graphic with a number: one verified number, large, with one sentence of context. Strong for credibility, as long as the number is real.
3. Provocative question: a question the target group asks itself. Strong as a hook.
4. Screen recording: a recording of a product, dashboard or configurator. Strong for software and anything that can be shown.
5. UGC: people showing a product in everyday life, usually produced by creators. More on this in the article UGC ads.
6. Carousel: several cards, for example process steps, products or before and after. Strong for explanation and choice.

Depending on the industry, there are also interior and location visuals (real estate), offer statics (e-commerce) or short clips from events (finance, B2B).

Testing structure and Monday review

Our standard for a Meta campaign:

1. No more than three campaigns per account, so that each gets enough data.
2. Broad audiences, Meta finds the people through the creative.
3. 4 to 6 concept creatives per ad set, each in three formats.
4. A backlog of prepared concepts, so that replacements are ready immediately.
5. Monday review: we read the week's numbers, decide on each creative, shift budget and top up the backlog.

In the Monday review, we ask three questions about each creative: Does it deliver the target metric below the target value? Does it have enough results for a decision? Does it show fatigue? This leads to three decisions: scale, keep running, replace.

A batch in practice

This is what a batch for one ad set can look like, here using a fictitious shop for coffee accessories as an example:

Every concept is produced in 1:1, 4:5 and 9:16. After the first week, usually one or two concepts remain that clearly lead. These get new hooks and lengths, and the weak ones are replaced by new concepts from the backlog.

When a creative wins

A creative wins when it delivers the campaign's target metric most cheaply, across enough results. Not when it has the most likes or the highest click-through rate.

A winner is often clearer than you would think. In the Funkhaus Living case, a new generation of creatives cut the cost per enquiry by 48 per cent, and the best creative came in at €4.70 per enquiry. In the Kommunalkredit case, the speaker creative got almost three times the clicks of locations.

Creative fatigue is the counterpart: frequency rises, click-through rate falls, cost per result rises over several days. Then you need new concepts, not cosmetic changes.

AI creatives: use and labelling

AI speeds up our testing. We use it for:

• Variants of hooks and copy, which a human selects and sharpens.
• Format adaptations, for example extending a visual from 1:1 to 9:16.
• Drafts and storyboards, before shooting.
• Voice-over in testing: in the Kommunalkredit case, we ran an AI voice-over test for an intro film, which the client approved after the first cut.

The rules for this, as of October 2026, not a substitute for legal advice:

1. Curated: every AI creative is reviewed by a human for brand, facts and tone.
2. Labelled: Article 50 of the EU AI Act (Regulation (EU) 2024/1689) requires transparency, among other things for realistic, artificially generated or manipulated images, videos and audio. Meta additionally labels certain AI content itself. We label AI creatives on request. We always comply with statutory labelling obligations (for example Art. 50 of the AI Act, such as for realistic AI depictions of people) and check this before uploading.
3. No fake people as real customers: an AI-generated "customer" with an invented experience is misleading advertising.
4. Check the rights: images, voices and music from AI tools only with clear usage rights.

Examples from our cases

The lesson from all four: the creative moves more than the budget. And the winner is rarely the one you would have backed beforehand.

What creative testing cannot do

Testing finds the best creative among those you test. If they are all weak, the least weak one wins. That is why every batch at ad.boutique starts with the target group, angles and offer, not with design. And testing needs budget: with very small budgets, we test fewer concepts, but for longer.

How we work at ad.boutique

At ad.boutique in Vienna and Zurich, creatives are made in the same team as the campaigns: strategists, designers, videographers and editors, supported by AI for more testing volume. Batches with 4 to 6 concepts, three formats and a backlog, evaluated in the Monday review, visible in the live dashboard in boutique.OS. More on the campaign side at Meta ads agency and on production at content production.