The modern blueprint for Meta ads settings: How to align your campaign structure with AI

If you are still using the “creme de la creme” Meta ads account structure from 2022, there is a good chance you’re throttling your campaigns performance.

For years, the best practice was to use lots of campaign segmentation, loading up campaigns with multiple interest groups, different lookalike percentages, and separate retargeting pools, all targeting different areas of the funnel.

However, Meta has completely overhauled its algorithm with the recent Andromeda update. 

Now, the best practice has been turned on its head. Before, combining everything into one campaign was seen as crazy, now, it can work wonders.

The big pivot: The Andromeda update

Throughout 2024 and 2025, Meta developed a new architecture for the algorithm, capable of handling the explosion of new ad variations generated by the new Advantage+ creative tools. This was called the Andromeda update.

Previously, the algorithm grouped 1000 ads together based on targeting settings, as it couldn’t process each individual ad in each auction. It was efficient, but heavily relied on targeting settings. Now, the algorithm can scan tens of millions of different ads, before deciding what ad the user should see.

Instead of looking at audience settings first, Meta now scans the creative assets, text, and hooks, before deciding which users are most likely to engage..

New Meta feature influx

With the introduction of so many new Meta ads settings across campaign, ad set, and ad level, the interface can feel a bit overwhelming, with some settings feeling like their function isn’t clearly explained, and new settings appearing, changing name, or moving location.

You may be looking at these Meta ads new updates and worry about letting the machines handle your campaigns, especially where its recommendations go against everything you’ve learned in the old Meta world. However, the goal isn’t to use every Meta AI ads feature, or go the other way, fighting the platform, and turning off all the AI settings out of fear. Finding the sweet spot between the two can be the key to success, giving the algorithm enough freedom whilst maintaining guardrails to ensure you’re still in control.

Audiences: Broad, Interests, Lookalikes, Retargeting

Since the Andromeda update, traditional audience targeting has started to die off, and the era of giving signals through creative is now the priority. Even when using open targeting, you can achieve great results, as the algorithm scans the signals through your creative, meaning Meta has a large canvas to find potential customers.

Advantage+ means that interests and lookalikes are now used as suggestions rather than strict guidelines, and can give your campaigns a boost when starting as, as the campaigns will know what users to look for. However, the biggest advantage comes from 1st party data. Using 1st party data as an audience signal, or turning it into lookalikes, can give your campaigns an effective kickstart.

The same fluid logic applies to Meta ads retargeting campaigns. Retargeting can still be effective, but having it in a separate campaign can be obsolete. Instead of using the old fragmented structures from 2022, your retargeting audience can sit within the same campaign as your cold audience ad set. This way, the algorithm has dynamic control over budget allocation based on live data and performance.

How the algorithm has rewritten the rules

To fully understand how Meta ads AI targeting works today, you need to fully embrace the saying that start popping up a few years ago: “Creative is the new targeting”. Because the Andromeda engine processes signals differently, the algorithm doesn't need to rely on profile checkboxes to find your optimal audience. Instead, Meta AI ads use your creative visuals and copy messaging as the primary factor.

If you want to sell a product specifically to parents, you no longer need to apply interests or demographic settings to search for them, instead, if you place an image or video of a child with relevant messaging, the AI ads Meta system is able to identify who this would be relevant for, and track who has stopped scrolling whilst coming across your ad. It then uses those behaviour signals to find other users likely to engage.

The modern blueprint

To help the algorithm work effectively, you need to focus on a structure that allows it to have enough freedom to find your optimal audience.This can mean grouping your budget into fewer, larger campaigns, rather than fragmenting them into smaller, precise pieces. When you set up your Meta account structure this way, you will no longer be forcing each of the ads to compete against each other in auctions, as well as allowing campaigns to gather enough data to exit the learning phase faster.

A solid Meta campaign structure can now be as simple as two campaigns. The core campaign being a scaling campaign, using 2 ad sets consisting of broad targeting (ideally using 1st party data as the audience signal), and retargeting in the other.

To keep the coals burning, you should be adding new creatives to the campaign often, to both give the algorithm more signals, and lowering the chance of audience fatigue. To optimally do this, pair the scaling campaign with a separate testing campaign. Following a testing plan like the one detailed here in our blog will give you a solid, modern, Meta ad campaign structure.

When to step in: The balancing act

Finding the right balance can be tricky as it varies from business to business. If your business's goal is to find a high volume of leads, sell widely used products or services, you can lean into Meta ads AI automation. In these cases, the algorithm thrives off wider targeting.

However, putting full trust into automation can sometimes lead to wasted budget, and sometimes, you need to step in. For example, we recently audited a paid social client’s account where the data showed that specific age ranges and demographics were spending consistently, but weren’t converting. The solution was to disable Advantage+ audiences to enforce manual exclusions. Whilst this meant that we couldn’t utilise audience signals, it meant spend wasn’t leaking to irrelevant users, and overall campaign performance improved.

How to avoid overcomplicating targeting

There’s nothing worse than checking an account and seeing CPAs skyrise for no reason. When performance sharply changes, the natural move for most marketers is to dive into the settings, and start tweaking them to force a fix. But with how the algorithm works now, keeping your Meta ads structure simple is usually what can help the most.

To check that your account is built to work with AI rather than fight against it:

• Switch your focus to creative: Creative volume and variation is key, where testing fresh styles, hooks, and messaging can be the key to an account's success. Let your creatives do the targeting for you.
• See where you can consolidate campaigns: Although you may not be able to combine everything into one campaign, finding campaigns with similar objectives and using them to create a single, larger pool can help the algorithm perform.
• Check your learning phases: Ad sets ideally need 50 conversions per week to exit the learning phase. Although this is optimistic for some businesses, making sure your account isn’t too fragmented to work towards this goal can give the algorithms an extra boost.

Keeping things simple can be the key to success in the world of Andromeda. Small targeting tweaks are no longer enough to fix a weak creative lineup. Focus on cleaning up your account setup, research into what the audience wants to see, and let the creative do the tinkering for you.