Meta Ads Campaign Structure in 2026: What Andromeda Changed

If your Meta ads account started behaving strangely toward the end of 2025, you’re not imagining it. A structure that used to deliver steady results suddenly felt random. Costs moved for no clear reason. A good ad lost its budget the following week.

The main reason is Andromeda, the new system Meta uses to match ads to people. So here’s my short answer. Build one campaign per offer, keep retargeting separate, and don’t hand everything to one big ad set. What’s stabilising results across my client accounts is a hybrid: one ad set with a pool of ads for Andromeda, plus a few single-ad ad sets you control.

What Andromeda changed

Around August to September 2025, Meta rolled out Andromeda, a rebuilt engine that decides which ad gets shown to which person. It wasn’t a settings change. It was more like swapping the engine while the car was still moving.

Meta has always been intent-based advertising. What changed is how much of the matching Meta now wants to do itself. Its recommendation for Sales, Leads and App campaigns is to go Advantage+ at every level, and to load one ad set with many ads, with suggestions as high as 20 to 30, each speaking to a different kind of person.

What actually happens in real accounts

I’ve used it across nearly all my client accounts. Two patterns keep showing up.

• The budget piles onto 2 or 3 ads. Out of 10 ads, a few take most of the spend. That’s fine while they perform.
• The winner doesn’t stay funded. An ad does well this week, and next week Meta pulls budget away from it to test the others. You can’t build a scaling plan on an ad that might be starved next week for reasons that have nothing to do with its results.

Here’s the thing. This isn’t a verdict that Andromeda doesn’t work. Across my clients it’s closer to 50/50. Some accounts do really well fully on Andromeda, others don’t, and you often can’t tell which in advance. So don’t get attached to a framework. Get attached to results.

The structure I use in 2026

For each offer, product line or category:

• One campaign for that offer. Different offers can each have their own. This isn’t “one campaign for the whole account”.
• An Andromeda ad set with a pool of ads, taking roughly half to 60% of the budget.
• A few single-ad ad sets with the rest, each testing one ad at a modest budget.
• Retargeting in its own campaign, running at the same time as prospecting from day one.

Let’s say, for example, you spend $100 a day and you have 10 ads. Put all 10 into the Andromeda ad set with $50 to $60 a day. Take 5 of those ads and run each one on its own at about $10 a day.

Single-ad ad sets are more stable because there’s no pool for Meta to reshuffle. If an ad runs at a certain CTR, CPM and cost per result, it tends to keep running at about that level. That stability is what lets you scale with confidence.

Then follow the numbers. If the Andromeda ad set stays stable, shift more budget into it over time. If it doesn’t, move weight to the single-ad side, which has already proven itself.

How to test new ads

Don’t throw 30 to 50 ads into one ad set and hope. Meta will still put most of the money behind one or two of them, often not the cheapest ones.

Test 3 to 5 ads at a time, give each a fair share of the budget for one to two weeks, then narrow down to the 1 or 2 winners. Retire what isn’t working, bring in a new ad, and keep cycling. Over time you build a bench of proven ads. Some accounts end up with 4 or 5, others with 10 to 15.

Get the free guide: 7 Things Every Meta Ad Needs to Sell →

A short PDF checklist built on a storytelling framework for ad creative and copy, so your ads have the pieces they need to sell.

Build ads for different people, not one ad for everyone

What goes into the pool matters as much as the structure. Instead of one ad that tries to please everyone, build a few ads that each speak to a different kind of buyer.

Imagine you sell kitchen appliances. A mother whose son just started baking might see an ad with a young man, framed as a gift. A father who loves family time might see an ad with the whole family baking together. Same product, same offer, different ad, and Meta matches each one to the people it fits.

Make each ad genuinely different: different picture, different angle, different first line. Not the same ad in a new colour.

Judge the pool, not each ad on its own

An ad built for a smaller group will get less delivery. An ad featuring a 60-year-old in an audience with few 60-year-olds will naturally see lower volume. That doesn’t make it a bad ad.

So before you pause something, ask whether it converts well for the people it does reach, not just whether it gets enough spend. Still watch CPM and CTR. Just read them across the whole set of ads, and give each change a week. We monitor daily, but we measure weekly.

A quick self-check

• Does each offer or product line have its own campaign?
• Is retargeting running in its own campaign, at the same time as prospecting?
• Do I have a stable single-ad ad set to fall back on if the Andromeda pool swings?
• Am I testing 3 to 5 ads at a time for one to two weeks, not 30 at once?
• Do my ads speak to different kinds of buyers, or are they the same ad in different colours?
• Am I pausing low-delivery ads only after checking how they convert?

If most answers are no, fix the structure before you touch the budget.

Frequently asked questions

Why did my Meta ads stop working in late 2025?

For many accounts it traces back to Andromeda, the new system Meta rolled out around August to September 2025 to match ads to people. Structures built for the old system, like many ad sets each testing a few ads, often stopped performing the same way.

What is the best Meta ads campaign structure in 2026?

One campaign per offer, with retargeting in its own campaign. Inside the offer campaign, a hybrid works well: one ad set with a pool of ads for Andromeda on about half the budget, plus a few single-ad ad sets you control. Then shift budget toward whichever side stays stable.

How many ads should I put in one Meta ad set?

For testing, 3 to 5 ads for one to two weeks, then narrow down to the 1 or 2 winners. Meta suggests far more, but in my experience it still puts most of the budget behind one or two ads, so a huge pool mostly adds noise.

Should I pause an ad that gets very little delivery?

Not straight away. An ad built for a smaller group of people will naturally get less spend. Check how well it converts for the people it reaches before you pause it.

Want me to look at your structure with you?

If your account has felt unpredictable since Andromeda, book a free 20-minute call with me. Just me, no sales team. We talk about where you are, what you want to fix, and whether working together makes sense.

If you want to go further, read more on the hybrid structure, why I like single-ad ad sets for testing, or why your best ad loses budget the next week.

Written by Jason Gan, a Meta ads strategist who has run Meta ads since 2010 and personally audited more than 1,000 ad accounts. Meta Certified Professional and Badged Meta Business Partner. Watch real account breakdowns on his YouTube channel.