ads-meta

A deep-analysis playbook for Meta advertising across Facebook, Instagram, and Threads, tuned to the 2026 Andromeda + GEM + Lattice delivery stack. It explains how Meta rebuilt delivery: Andromeda is an ad-retrieval ranking model that filters the candidate creative set before the auction; GEM (Generative Embedding Model) embeds creative content directly into the targeting space, making "creative is the new targeting" a mechanical reality; and Lattice is a sequence-aware optimizer that ranks ads from user-action sequences. The practical consequence is that creative diversity is the top performance lever, because near-identical creatives above a measured similarity threshold get retrieval suppression, so 100 minor variations perform no better than 10 genuinely distinct ones. The skill supplies a creative-as-targeting scoring rubric (concept, format, visual, hook, and headline diversity, each scored 0-2 for a 0-10 total that maps to LOW/MEDIUM/HIGH Entity-ID clustering risk) and an Entity-ID clustering predictor with pre-launch heuristics on visual, headline, body, and video-hook fingerprints that produces a creative-cluster-risk deliverable. It also includes a MAPI v25 ASC/AAC deprecation detector explaining how Advantage Shopping and Advantage App Campaigns fold into standard Sales/Leads/App objectives where ASC behavior becomes the default, plus the audited ASC defaults (catalog connection, an existing-customer cap of 10-25%, Advantage+ Audience, and Advantage+ Creative). The audit process runs 50 weighted checks and produces a Meta Ads Health Score from 0-100, drawing on reference files for the full audit, benchmarks, and scoring. Weighted areas include Pixel/CAPI health (30%): pixel firing, Conversions API, event deduplication, EMQ at least 8.0, standard events, AEM, and domain verification; Creative (30%): format and creative counts, fatigue detection, video lengths, copy limits, and refresh cadence; Account Structure (20%): CBO vs ABO, consolidation, learning phase, and audience overlap; and Audience & Targeting (20%). Use it whenever auditing or planning Meta campaigns.