In this article
What Andromeda actually does
Meta describes retrieval as the first stage of its multistage ads recommendation system. Andromeda selects a smaller relevant candidate set from a much larger pool, before later ranking stages. [1] It is a system explanation, not a setting an advertiser directly toggles.
Keep the date precise. The primary engineering article is from December 2024. This guide’s 2026 review does not mean a new global Andromeda launch happened on the day this page was updated. Distinguishing publication, product evolution and account observation prevents misleading rollout narratives.
How it differs from GEM and Advantage+ features
Meta’s October 2026 GEM article describes a central recommendation foundation model with learning across surfaces and objectives. [2] Advantage+ is the name advertisers encounter across automation products and controls. These terms describe different layers and should not be used as synonyms.
When a client asks what changed, identify the actual evidence: a documented system explanation, a newly available account feature or a change in account metrics. Each can be relevant, but only the last category establishes what happened in that client’s campaigns.
Translate the system discussion into a useful creative review
List the buying problem, promise, evidence and destination for each concept actually running. Check whether assets communicate different reasons to buy or merely repeat a headline in different colors. Use genuine product and customer evidence for any claims.
A review can recommend a new concept without claiming the model requires a universal number of creatives. The useful question is whether the next asset addresses a real gap and can be evaluated against an appropriate business outcome. More production is not automatically better experimentation.
Check the measured signal and business outcome
Verify which action the campaign optimizes and how it is recorded. Compare it with the outcome the business values. A campaign can deliver cheap superficial actions while failing to generate qualified enquiries. Conversely, incomplete tracking can make real business activity look weaker than it is.
Keep native diagnostic evidence, website events and backend records distinct. If qualification data is unavailable, say so. Do not infer that a particular percentage of server-side events proves a healthy implementation; investigate actual event definitions and delivery with the implementation owner.
Use an account-specific comparison
Compare complete periods with matching conversion definitions. Break down spend and outcomes by campaign and market, then examine the largest contributors. Record changes in offer, website, tracking and advertising settings during the same period.
For a hypothetical account, a blended cost rise may come from allocating more spend to a higher-cost market. That can happen even when within-market performance remains stable. Reconstructing the allocation gives a more actionable explanation than assigning the whole change to an unnamed algorithm effect.
Choose the next action with evidence
In 1ClickReport, select the client, connect Meta and ask the Analyst to investigate the largest change with current data. Use client memory for confirmed goals and earlier decisions, and refresh performance evidence before making a recommendation.
The final brief should identify one test, its supporting observation and what would make the result inconclusive. A creative, destination or measurement change may be justified, but its outcome needs verification. This guide does not claim anonymous recovery cases, a guaranteed timeline or results the source cannot substantiate.
A first task to try
Review this client’s Meta account using current connected evidence. Identify the largest campaign and market contributors to the change, then inspect the business outcome and creative-message gaps. Separate Andromeda context from proven causes and propose one reviewable next test.Questions readers ask
Can I enable Andromeda myself?
It is Meta’s retrieval system rather than an advertiser-facing campaign toggle.
Will a fixed creative count guarantee better ROAS?
No. Evaluate message, delivery, data quality and business outcomes in the actual account.
Sources and editorial method
Research checked October 26, 2026. Platform facts link to primary sources below. The worksheets and operating examples are our editorial recommendations; hypothetical examples are not customer results. This article is published by 1ClickReport.