Meta Ads Deep Analysis

Andromeda + GEM + Lattice (2026)

Meta's delivery stack was rebuilt across three releases:

• Andromeda (Oct 2025) — ad-retrieval ranking model with 10,000× more model capacity than the previous funnel (Meta Engineering, Dec 2024). Filters the candidate creative set before the auction layer ever sees it.
• GEM (Generative Embedding Model, late 2025) — replaces the feature pipeline. Creative content embeds directly into the targeting space, which is why "creative is the new targeting" is now mechanical truth not slogan.
• Lattice (rolled out late 2025 / early 2026) — sequence-aware optimizer on top of GEM that uses user-action sequences to rank candidate ads.

Net effect: creative diversity is now the #1 performance lever. Ads with Similarity Score >60% (per Confect's measured threshold) get retrieval suppression — the algorithm clusters near-identical creatives and silently limits their delivery. 100 minor variations perform no better than 10 genuinely distinct ones. Prioritize concept / angle / format diversity over variant volume.

Creative-as-targeting scoring rubric

When auditing a creative library against Andromeda's retrieval logic, score across these 5 axes (each 0-2, total 0-10):

Score 8-10 = LOW Entity-ID clustering risk. Score 4-7 = MEDIUM risk (some suppression likely). Score 0-3 = HIGH risk (significant retrieval ticket loss).

Entity-ID Clustering Predictor (pre-launch)

Before launch, predict which creatives Meta will cluster. Cluster-mates share retrieval tickets — only one wins per impression opportunity.

Predictor heuristics (apply to every pair of creatives in the launch set):

1. Visual fingerprint — same product hero, same model, same backdrop, same lighting → likely cluster. Different products or different visual identities → likely not a cluster.
2. Headline fingerprint — same first 4 tokens → likely cluster (e.g. "Save 30% on" + "Save 30% off" + "Save 30% — limited time").
3. Body copy fingerprint — same opening sentence, same CTA verb → likely cluster regardless of middle-body differences.
4. Video hook fingerprint — same 0-3s shot, same voiceover pattern → likely cluster even if the rest of the video diverges.
5. Format mismatch wins — if pair is (static + video) AND visual fingerprint differs, they are not clustered. Crossing format AND visual is a strong diversity signal.

Output: produce a creative-cluster-risk.md deliverable that groups the launch set into predicted clusters, recommends which creative in each cluster should ship and which should be cut or rebuilt, and reports the final pre- launch diversity score (target ≥8/10).

MAPI v25 ASC/AAC Deprecation Detector

Meta Marketing API v25 deprecates the explicit Advantage Shopping Campaigns (ASC) and Advantage App Campaigns (AAC) creation paths — those campaign types are folded into standard Sales / Leads / App objectives where ASC behavior becomes the default configuration. Detection:

• If the account uses MAPI v23 or earlier: ASC/AAC API endpoints will return deprecation warnings before the v25 cutover. Capture and flag them.
• If the account uses MAPI v25+: confirm that previously-ASC campaigns have been migrated to the new objective-default model with the equivalent catalog + budget + existing-customer cap settings preserved.
• If creating new campaigns: use the Sales / Leads / App objective + ASC defaults rather than the legacy ASC/AAC endpoints.

ASC defaults for Sales / Leads / App (2026 behavior)

When Sales / Leads / App objectives are selected, ASC behaviors are now the default. Audit confirms:

• Catalog connection (Sales): product catalog linked and feed health green
• Existing customer cap (Sales): set to 10-25% (default may be too high for high-LTV brands)
• Advantage+ Audience (all three objectives): on by default; only override with manual interest stacks for highly restricted categories
• Advantage+ Creative (all three): text / brightness / music enhancements on by default; if your brand-safety policy requires off, document the exception per ad set

Process

1. Collect Meta Ads data (Ads Manager export, Events Manager screenshot, EMQ scores)
2. Read ads/references/meta-audit.md for full 50-check audit
3. Read ads/references/benchmarks.md for Meta-specific benchmarks
4. Read ads/references/scoring-system.md for weighted scoring
5. Evaluate all applicable checks as PASS, WARNING, or FAIL
6. Calculate Meta Ads Health Score (0-100)
7. Generate findings report with action plan

What to Analyze

Pixel / CAPI Health (30% weight)

• Meta Pixel installed and firing on all pages
• Conversions API (CAPI) active (30-40% data loss without it post-iOS 14.5)
• Event deduplication configured (event_id matching, ≥90% dedup rate)
• Event Match Quality (EMQ) ≥8.0 for Purchase event
• All standard events configured (ViewContent, AddToCart, Purchase, Lead)
• Custom conversions created for non-standard events
• Aggregated Event Measurement (AEM) configured for iOS
• Domain verification completed
• Server-side events include customer_information parameters
• Pixel fires with correct currency and value parameters

Creative (30% weight)

• ≥3 creative formats active (image, video, carousel, collection)
• ≥5 creatives per ad set (Meta recommendation)
• Creative fatigue detection: CTR drop >20% over 14 days = FAIL
• Video creative: 15s max for Stories/Reels, 30s max for Feed
• UGC/testimonial creative tested
• Dynamic Creative Optimization (DCO) tested
• Ad copy: headline under 40 chars, primary text under 125 chars
• Creative refresh cadence: every 2-4 weeks for high-spend

Account Structure (20% weight)

• Campaign Budget Optimization (CBO) vs Ad Set Budget (ABO) intentional
• Campaign consolidation: 1-3 campaigns total recommended
• Learning phase health: <30% ad sets in "Learning Limited" (FAIL >50%)
• Budget per ad set: ≥5x target CPA (minimum for learning phase exit)
• Ad set audience overlap <30% (Audience Overlap tool)
• Campaign naming conventions consistent and descriptive
• Advantage+ Sales Campaigns active for e-commerce
• Simplified campaign structure: 1-3 campaigns total (fewer, larger ad sets preferred)

Audience & Targeting (20% weight)

• Prospecting frequency (7-day): <3.0 (WARNING 3-5, FAIL >5)
• Retargeting frequency (7-day): <8.0 (WARNING 8-12, FAIL >12)
• Custom Audiences: website visitors, customer lists, engagement
• Lookalike Audiences: multiple seed sizes tested (1%, 3%, 5%)
• Advantage+ Audience tested vs manual targeting
• Interest targeting: broad enough for algorithm optimization
• Exclusions: purchasers excluded from prospecting, overlap managed
• Location targeting reviewed for relevance

Advantage+ Assessment

If Advantage+ features are in use:

• Advantage+ Sales Campaigns: catalog connected, existing customer cap set
• Advantage+ Audience: performance vs manual audience compared
• Advantage+ Creative: e