Meta Campaign Structure in Andromeda Era

By Veronica Ruiz - Wednesday June 24, 2026 Share (Twitter) WhatsApp Summarize ChatGPT Perplexity Grok Google AI A conversation I have been having lately, when auditing Paid Social Ads campaigns: “The structure looks great! … for 2024. Unfortunately, this won’t work in 2026.” “Why not?”

Those who are not in the industry suddenly think I’ve gone a bit too poetic when I mention….. Andromeda! So I quickly have to clarify that this is Meta’s new retrieval-based, AI-powered ad retrieval and ranking system. It replaced the previous ad infrastructure and represents the most significant change to how Meta serves ads in over a decade.

For those who like interesting facts: Meta chose the name as a poetic nod to the system’s sheer scale. Just as the Andromeda galaxy contains over a trillion stars and is constantly expanding, Meta’s algorithm was designed to search through tens of millions of potential ad-to-user matches in real time. So, actually, the answer is poetic after all.

Above: A self created diagram From Micro-Planets to Galaxy Signal Design Until this galaxy revolution, Paid Social managers have been structuring their accounts based on heavy segmentation by audience and placement with separate ad sets for interests, lookalikes, and retargeting. And of course, with only up to 5-6 ads per ad set.

But Andromeda is significantly more advanced. It consolidates signals across audiences, placements, and behaviours to find conversion efficiency at scale. It needs “a good chunk” of data and “a good chunk” of creatives (so yes, we will need more than 5 ads per ad set).

Because in this large galaxy, our job is not to “control” performance, but to design systems that generate strong, clean signals for it. Consolidation: The way to create strong, clean signals Consolidation is key to creating these strong, clean signals.

The algorithm is effectively searching through millions of potential ad-to-user combinations, so you want your data concentrated enough for Andromeda to detect patterns clearly. This enables faster learning and smoother scaling.

So instead of spreading the budget thin across multiple campaigns and ad sets, Meta increasingly rewards: Fewer campaigns Fewer ad sets More budget per learning system Broader targeting inputs Campaign Structure Should Mirror Business Goals, Not Audiences The first step in effective consolidation is structuring campaigns around business goals – not audience segments.

This is the most scalable approach as it unifies signals towards the final business goal: So, instead of: “TOF / MOF / BOF” You should think: Acquisition vs Retention Lead vs Sales Geographic or operational constraints Using Broad as Expanding Universe If Meta needs consolidated data, broader targeting becomes the natural extension of that strategy.

We’re moving away from manual control, granular segmentation, and rigid funnel logic. Like it or not, Meta now has far more behavioural and predictive data on users than we can realistically process. This makes strict audience-based targeting increasingly obsolete. A strong and diverse creative system is becoming the new targeting layer.

Creatives, the New Targeting Layer Before Andromeda, audiences determined delivery. However, Andromeda now uses creative signals to determine which users should see an ad. The goal of this retrieval engine is to create a “shortlist” of strong ads – not one universal winner, but a set of highly relevant variations that resonate differently across users.

It is your hook and your message that determine audience qualification. And this is where creativity and diversity becomes important: a “winning” ad for one person may be completely irrelevant for another, even if both share the same declared interest (for example, oily skin beauty products).

Andromeda’s objective is to personalise delivery so the right user sees the right variation. It’s no longer about quantity alone; it’s about meaningful variety. The More Variety, the Higher the Chances of Finding Winners Andromeda learns not only from clicks or conversions, but also from engagement patterns and interaction behaviour.

Small tweaks are no longer enough. Changing copy. Swapping backgrounds. Adjusting overlays. Reusing the same creator in slightly different shots. Meta increasingly interprets this as the same ad. We are past the “find 7 differences” era.

“Different” now means actually different: Different concepts and angles Different narrative frameworks (problem/solution, pain point, AIDA, testimonial-led, contrarian takes) Different formats (UGC, static, video, carousel) Different personas and funnel stages Scaling Comes From Structured Creative Diversity Talking about tapping into different personas, this is also one of the most reliable ways to scale.

Different people buy for different reasons (even if both have shown an interest in “interior decor”). The more motivators or hooks you put in front of