The "Andromeda" update has shifted Meta from an "Identification Engine" to a "Predictive Intent Engine." Manually stacking interests now actively punishes your CPA by restricting algorithmic liquidity.
January 7, 2026
For a decade, the logic of Meta advertising was "Identification": the advertiser manually identified the user using interest tags. The deployment of the Andromeda Protocol has transitioned the platform to a "Predictive Intent Engine."
The algorithm no longer seeks "Mothers aged 25-34." It utilizes Large Action Models (LAMs) to seek "Users currently exhibiting maternal purchasing intent patterns." This intent is predicted in real-time based on micro-interactions—scroll speed, dwell time, and session context.
"Liquidity" refers to the freedom given to the algorithm to spend budget where it sees the highest efficiency.
Old-school tactics like "Interest Stacking" (combining 10+ narrow interests) or splitting placements (Feed vs. Stories) create rigid constraints. In the Andromeda era, these constraints act as friction. They force the algorithm to bypass the most efficient inventory because it doesn't fit your manual rules.
The fear of "Broad Targeting" is based on an outdated definition. In 2020, "Broad" meant spraying ads to everyone. In 2026, "Broad" means "Unconstrained Liquidity."
When you select an audience of "Men, 18-65+, US," you are not targeting all men. You are telling Andromeda: "Scan this entire pool, find the people with the highest Estimated Action Rate (EAR) for my specific creative, and serve the ad to them."
Restricting this pool with interest tags artificially limits the AI's ability to find lower-cost conversions outside your defined box.
This was the theory. Now get the schematic.
We have mapped the full "Andromeda Protocol" and the specific account structures required to maximize liquidity.
