Meta Andromeda does not care about your campaign structure

Meta Andromeda does not care about your campaign structure. It never did.

The three-layer system running under your Meta account, Lattice for intent prediction, Andromeda for creative matching, GEM for auction dynamics, operates horizontally across everything you have built. Campaign boundaries, ad set budgets, audience segments: all of it is input, not instruction.

Andromeda finds the best creative for the person in the auction regardless of which campaign that creative lives in. Budget liquidity means Meta decides which ads absorb spend. You decide how much to invest. Meta decides distribution.

Most advertisers are still auditing Meta accounts by looking at ad sets and budgets first. That is reading the wrong layer.

Why structure stopped mattering

The 2020 playbook treated Meta like a campaign architecture problem. Build the right account tree, segment audiences properly, ringfence budgets, monitor CPM by ad set. The thinking was that good structure compounded into good performance.

That was true when Meta's auction was less automated. Today, the system explicitly overrides most of those decisions. If your "low-funnel retargeting" ad set has a creative that resonates with cold audiences, Andromeda will serve it to cold audiences. Your ad set name does not change the model's mind.

The practical consequence: running 5 ads in a campaign and monitoring which one wins is not creative testing. It is handing a very small sample to a system designed for scale, then making budget decisions based on what the system was not given enough to learn from.

What actually works now

The practitioners who figured this out earlier are running 50 to 100 ads per month per account. Not because they have better creative intuition. Because Andromeda needs surface area to find winners that no human could predict upfront.

The pipeline shape that works:

• A documented creative brief refreshed monthly, with 6 to 10 angles to test per month
• Production capacity to ship 8 to 15 ad variations per week (static, video, UGC, native, founder voice)
• A library of historical winners with annotations on what worked and why
• Weekly review of which creatives Andromeda fed budget to, what the new winners share, and what is now plateauing
• Retired creatives that get rebuilt with different framings rather than disappearing

The account management skill that mattered in 2020 was structure. The skill that matters now is creative pipeline velocity.

What to actually do

If your Meta account is performing flat and the conversation in your team is about budget allocation between ad sets, you are diagnosing at the wrong layer.

• Audit your monthly creative output. Anything under 30 new ads per month per active campaign is starving Andromeda.
• Stop calling 5-variant ad-set tests "creative testing". Five is not enough variance for a system optimising over millions of impressions.
• Move budget decisions up a level. Stop adjusting ad-set budgets weekly. Pick a campaign-level spend, trust the budget-liquidity layer, focus your team's time on producing more creative.
• Invest in a creative ops function, in-house or agency. Most accounts spending $30k+/month on Meta justify a dedicated production line. Most do not have one.
• Track creative-cycle metrics, not just performance metrics. How long from brief to live? How many concepts per month? What proportion of new creatives become winners?

The accounts that grow on Meta in 2026 are the ones treating creative as a manufacturing function, not a periodic project. Andromeda rewards velocity. Structure is just plumbing.

If your current Meta setup is running flat and you want a view on whether the bottleneck is structure or creative, book a free audit and we will tell you which one is actually capping the account.