Meta Andromeda: why your creative is now your targeting

What Andromeda is

When someone opens Facebook or Instagram, Meta has to choose which ad to show them out of millions. It does this in stages. The first stage is called retrieval: it narrows tens of millions of possible ads down to a few thousand. Later stages rank those few thousand and pick the winner.

Andromeda is Meta's system for that first stage. Meta's engineers described it in a December 2024 post. They reported a 6% improvement in recall, which means finding more of the right ads for each person, and an 8% improvement in ads quality on selected segments.

The same post explains why Meta built it. Generative AI tools mean advertisers are making far more ads. Meta said over a million advertisers used its AI tools to create more than 15 million ads in a single month. Andromeda was designed to handle that growing pile of creative and still find the right ad for each person.

What Meta has said since

Meta keeps investing in this. In its results for the end of 2025, reported by PPC Land in January 2026, Meta said it had nearly tripled Andromeda's compute efficiency by running it on more types of chips. It also reported an average 22% improvement in return on ad spend from Advantage+ across 2025. Advantage+ is Meta's name for its automated campaign settings for audience, placements, budget and creative. Automated campaigns are now the default for sales, leads and app objectives, though manual controls are still there.

Why your creative is now your targeting

Meta has not used the phrase "creative is targeting". That is how many advertisers, including me, sum up what follows from the way the system works.

Here is the logic. With Advantage+ audience, you give Meta a broad audience and let it find the buyers. Andromeda then decides which ads are worth considering for each person. The main thing it has to go on is the ad itself: the image or video, the words, the product, the person in it, the offer. An ad showing a mum with a toddler in a Limassol park will reach different people than an ad showing a student in a Thessaloniki café, even with the same audience settings.

So if you run five ads that look almost the same, you have really given Meta one idea. It has one way to reach people. If you run three ads that are clearly different, you have given it three ways in.

What Meta's help pages say about ad volume

This does not mean "make as many ads as possible". Meta's own page on managing ad volume says the opposite: when you run too many ads at once, each one is shown less often, fewer of them exit the learning phase, and performance can get worse. Meta suggests:

• Combine similar ad sets, so learning is pooled.
• Use fewer ads per ad set, but keep the creative inside them diverse. One ad can hold up to ten creative assets.
• Test several text versions inside one ad, rather than making a new ad for each.

Meta's page on learning limited ad sets lists the usual causes: a small audience, a low budget, an event that happens rarely, or too many ads running at the same time.

The message is fewer, broader ad sets, with fewer but more different ads inside them.

A creative test that works on a small budget

This is the approach I use for local businesses. It keeps the number of ads low and the ideas far apart.

1. One campaign, one ad set. Broad audience, Advantage+ placements, one clear goal such as purchases or leads. Make sure the pixel and conversion events work first.
2. Three to four ideas, not twenty ads. Each idea should differ in something big: the reason to buy, the type of customer, or the format. For example, a price-and-delivery angle, a "how it is made" video, and a customer review.
3. Variations inside each ad. Add a few headlines and texts to each ad and let Meta mix them. That covers the small differences without splitting your budget.
4. Leave it alone for a week. Meta's learning phase page says ad sets usually exit learning after about 50 results in the week after the last significant edit. Editing during learning resets it.
5. Judge on cost per result, not likes. Look at cost per purchase or lead. Keep the winners running.
6. Add one new idea every couple of weeks. Do not replace everything at once.

When to refresh creative

Meta now tells you when an ad is tired. Its page on creative fatigue explains two statuses in the Delivery column. "Creative limited" means your cost per result is higher than your past ads, but less than double. "Creative fatigue" means it is double or more. Meta's advice is to create a new ad with an image or video that is materially different, and notes that keeping the original ad running may still give better results than pausing it.

What this means for businesses in Cyprus and Greece

Local audiences are small. In Cyprus especially, the same people see your ads again and again, so fatigue comes faster than in big markets. That makes variety in your creative more important, not less.

The good news is that local businesses have things big brands do not: real staff, real shops, real customers and real places. A phone video of the owner explaining a product in Greek can be a very different "idea" from a polished product photo, and Meta will treat it that way. Language is also a signal. An ad in Greek and one in English will tend to reach different people, which is useful in Cyprus where both are common.

What does not work any more is building ten ad sets by interest and age with the same picture in each. Put that effort into the creative instead.

Questions people ask

Want help with this? Book a free 30-minute call.

Sources

1. Meta Andromeda: Supercharging Advantage+ automation with the next-gen personalized ads retrieval engine, Engineering at Meta, December 2024
2. Meta's ad business hits record $58B as AI drives conversion gains, PPC Land, January 2026
3. About the learning phase, Meta Business Help Centre
4. About managing ad volume, Meta Business Help Centre
5. About learning limited, Meta Business Help Centre
6. About creative fatigue recommendations in Meta Ads Manager, Meta Business Help Centre