The old Meta Ads playbook is broken. Not slightly less effective — actively counterproductive in many cases. The campaign structures, audience segmentation approaches, and creative testing methods that drove results from 2019 to 2023 are now frequently the reason CPLs are rising. The reason is Andromeda. Meta's Andromeda is the AI retrieval and ranking system underneath Facebook and Instagram advertising. It replaced traditional interest-based audience matching with something closer to what Google's Smart Bidding does: a model that identifies converting users from signals no manual targeting system would have specified. It rolled out progressively from Meta Connect 2024 and reached full deployment in 2026. The practical consequence: your audience targeting selections now matter far less than they used to. Andromeda finds the audience. Your job is to give it the clearest possible signal of what a good conversion looks like — and get out of its way.
What Meta Andromeda Actually Is
Andromeda is Meta's AI-powered ad retrieval system — a deep learning architecture that operates at the ad-serving layer to match ads with users more precisely than any rule-based targeting system could. Traditional Meta ad targeting worked by advertisers selecting demographic segments, interest categories, and lookalike audiences, then Meta serving those ads to matching profiles. Andromeda inverts this logic: advertisers define their objective and provide high-quality creative signals, then Andromeda's model dynamically identifies which users across Meta's entire audience pool are most likely to convert — including users that no manual interest segment would have selected. The system processes billions of signals per second: user interaction patterns, content engagement history, purchase intent signals from Meta's off-platform data partnerships, creative performance history, and landing page relevance scores. Andromeda then ranks every ad-user pairing in real time and serves the combinations most likely to drive the advertiser's objective. The key implication: your audience targeting selections are now much less important than they used to be. Andromeda will find your audience — your job is to give it the clearest possible signal of what a good conversion looks like.
- Andromeda is a retrieval and ranking AI, not just an audience targeting upgrade — it fundamentally changes the ad-serving architecture
- The system evaluates billions of ad-user signals in real time to make serving decisions that no manual targeting system could replicate
- Creative quality, landing page relevance, and conversion signal clarity are now the primary levers advertisers control
- Broad audience targeting combined with high-quality creative now outperforms tight manual interest targeting in most verticals
- Andromeda learns from your conversion events — cleaner pixel data and server-side Conversions API tracking accelerates its optimisation
How Andromeda Scores Your Ads
Understanding how Andromeda evaluates ads is the fastest path to improving Meta Ads performance in 2026. The system applies a multi-factor quality score to every ad before it enters the auction — conceptually similar to Google Ads Quality Score but far more sophisticated. Andromeda's ad scoring evaluates four primary dimensions. First, creative quality signals: visual clarity, text-to-image ratio, hook strength in the first 3 seconds of video, and whether the creative matches the messaging on the destination landing page. Second, landing page relevance: Andromeda's crawler evaluates your landing page for content relevance to the ad, page speed, mobile experience, and the presence of a clear conversion mechanism. Third, historical performance signals: the ad account's track record of delivering genuine conversions (not just clicks), audience quality from previous campaigns, and the learning phase completion rate across active campaigns. Fourth, conversion signal quality: the reliability and completeness of your pixel events or Conversions API data, the volume of purchase or lead events in your account history, and whether your reported conversions match Meta's predicted conversion models. Advertisers with clean Conversions API setups, high-quality creative, and relevant landing pages receive dramatically lower CPMs from Andromeda because the system predicts they will deliver genuine value to users — and genuine revenue to Meta.
- Creative quality is now directly scored by Andromeda — low-quality visuals receive higher CPMs regardless of audience targeting
- Landing page relevance to ad copy is evaluated and factored into Andromeda's quality score — message match matters
- Conversion signal quality (CAPI completeness, event match quality) directly improves Andromeda's ability to find converters
- Accounts with learning phase completion history perform better — avoid frequent campaign restarts
- Creative diversity within ad sets provides Andromeda more signals to identify what resonates with converting audiences
Campaign Structure Changes for the Andromeda Era
The optimal Meta Ads campaign structure has changed significantly under Andromeda. The tactics that dominated 2022–2024 — hyper-segmented ad sets, tight custom audiences, manual interest stacking — now frequently underperform consolidated, broad-targeting structures that give Andromeda maximum audience exploration freedom. The recommended structure for 2026 is what practitioners call the 'Consolidated Creative Testing' model: one campaign per objective, one or two broad ad sets with minimal audience restrictions, and three to five distinct creative variations per ad set. This structure works because it gives Andromeda a large enough audience pool to identify patterns, enough creative variations to run internal A/B tests, and clear objective signals to optimise toward. Specific structural changes to make: replace interest-based ad sets with broad targeting or Advantage+ Audiences and let Andromeda do the targeting work; consolidate any ad sets spending under $50/day into single consolidated ad sets to exit the learning phase faster; use campaign budget optimisation (CBO) rather than ad set budget optimisation so Andromeda can dynamically allocate budget to the highest-performing creative-audience combinations; and implement a 7-day click attribution window aligned with your actual sales cycle length.
- Consolidate to 1–2 broad ad sets per campaign rather than multiple segmented interest-based audiences
- Use Advantage+ Audiences (formerly Broad Targeting Expansion) to give Andromeda full audience exploration freedom
- Campaign Budget Optimisation (CBO) outperforms ad set budgets under Andromeda's architecture
- Minimum 3–5 creative variations per ad set; Andromeda needs variety to identify top-performing signals
- Avoid frequent creative rotation or campaign restarts — learning phase history is a valuable Andromeda input
The Creative Strategy Andromeda Rewards
If audience targeting is now Andromeda's job, creative strategy is definitively the advertiser's primary competitive advantage. Andromeda's quality scoring means that creative quality directly affects your CPM — high-quality creative lowers auction costs, while low-quality creative raises them. In 2026, 'quality' in Andromeda's scoring means several specific things. For static ads: a clear visual hierarchy that communicates the core value proposition without requiring text, strong contrast between foreground and background, minimal text overlay (under 20% of image area), and an image that directly represents the product or service being advertised. For video ads: a hook that delivers the core value proposition or pattern interrupt within the first 3 seconds, native-feeling production quality (raw phone footage frequently outperforms polished agency production on mobile feeds), captions for sound-off viewing, and a clear call-to-action before the 15-second mark. For lead generation objectives specifically: creative that pre-qualifies the viewer by stating who the offer is for ('For B2B companies spending $5K+ on Google Ads'), showing social proof before the CTA, and matching the visual language of the landing page the ad sends traffic to. The creative-to-landing-page consistency signal is one Andromeda weights more heavily than most advertisers realise — disconnected visuals between ad and landing page trigger quality score penalties that inflate CPMs.
- Hook strength in first 3 seconds is the single highest-weighted creative variable in Andromeda's video scoring
- Creative-to-landing-page visual and message consistency is a distinct Andromeda quality signal — mismatches penalise CPMs
- Raw, native-feeling mobile content frequently outperforms polished production on Instagram and Facebook feeds
- Pre-qualification language in creative ('For businesses spending $2K+/month on ads') improves lead quality scores
- Test at minimum 3 creative angles per product — problem-led, benefit-led, and social proof-led — to find what Andromeda optimises toward
Tracking and Attribution Under Andromeda
Andromeda's AI optimisation is only as good as the conversion signals it receives. Meta's Conversions API (CAPI) is now the baseline requirement for Andromeda to perform — not an optional enhancement. Advertisers still relying solely on browser pixel tracking are providing Andromeda incomplete conversion data, causing the system to under-optimise, over-spend on non-converting audiences, and deliver inflated CPLs. The CAPI feeds server-side conversion events to Meta's systems with higher match quality, better deduplication, and iOS-independent attribution. Combined with the browser pixel for redundancy, server-side tracking typically recovers 15–30% of conversions that browser tracking misses entirely due to ad blockers, iOS restrictions, and browser cookie limitations. For Andromeda optimisation specifically, Event Match Quality (EMQ) — Meta's internal measure of how accurately your conversion events match to real Facebook users — should be above 7.0 for optimal Andromeda learning. Check your EMQ in Meta Events Manager under 'Web Events' for each event. If your lead or purchase EMQ is below 6.0, passing additional user data parameters (email hash, phone hash, first name, last name) via CAPI will significantly improve Andromeda's conversion signal quality and accelerate campaign optimisation.
- Conversions API (server-side tracking) is now a baseline requirement for Andromeda optimisation — browser pixel alone is insufficient
- Event Match Quality (EMQ) above 7.0 in Meta Events Manager indicates strong Andromeda learning signal quality
- Passing hashed email and phone data via CAPI improves EMQ and accelerates audience finding under Andromeda
- CAPI + pixel redundancy setup typically recovers 15–30% of conversions missed by browser-only tracking
- Higher EMQ directly reduces CPL — Andromeda can find converters more efficiently when it has complete signal data
The businesses that figure this out first won't just have better CPLs — they'll build a performance gap that's hard to close. Andromeda learns from conversion signals, and the more clean signal you feed it, the better it gets at finding your buyers. That compounds. The to-do list is shorter than most agencies make it seem: get your CAPI connected and your Event Match Quality above 7.0, consolidate your ad sets so Andromeda has room to work, stop restarting campaigns that are in the learning phase, and invest seriously in creative testing beyond 'make three ads and see what happens.' None of this is mysterious. It's just the work that most accounts haven't done yet.
Frequently Asked Questions
What is Meta Andromeda and how is it different from previous Meta ad targeting?
Meta Andromeda is an AI-powered ad retrieval and ranking system that replaced Meta's traditional interest-based audience targeting with a semantic, entity-based matching architecture. Rather than advertisers selecting who sees their ads based on interest categories, Andromeda dynamically identifies the users most likely to convert by processing billions of behavioural and contextual signals. The key difference: advertiser audience selections now matter much less than creative quality and conversion signal clarity, because Andromeda finds the audience automatically.
How should I change my Facebook Ads campaign structure for Andromeda?
The key structural changes for Andromeda: (1) consolidate to 1–2 broad ad sets per campaign rather than multiple segmented interest audiences, (2) use Advantage+ Audiences to give Andromeda maximum targeting freedom, (3) use Campaign Budget Optimisation (CBO) rather than ad set budgets, (4) maintain 3–5 creative variations per ad set, and (5) avoid frequent campaign restarts that disrupt the learning phase. Andromeda performs best with consolidated structures that provide large audience pools and diverse creative signals.
Does Andromeda mean I should stop using custom audiences and lookalikes?
Not entirely, but their role has changed. Custom audiences (website visitors, CRM lists) remain valuable as exclusion audiences (preventing ads from serving to existing customers) and for retargeting warm audiences. Lookalike audiences are less necessary than before because Andromeda's broad-match AI targeting typically outperforms manually constructed lookalikes at scale. For prospecting campaigns, broad targeting with Advantage+ Audiences now generally outperforms 1–5% lookalikes in most verticals.
How does server-side tracking (CAPI) improve Andromeda performance?
Andromeda's AI optimises based on the conversion signals it receives. Server-side Conversions API (CAPI) sends higher-quality, more complete conversion data to Meta — recovering events missed by browser pixel due to iOS restrictions, ad blockers, and cookie limitations. Better conversion data means Andromeda can more accurately identify what type of user converts for your offer, reducing CPL over time. Accounts with high Event Match Quality scores (above 7.0 in Meta Events Manager) consistently outperform low-EMQ accounts in Andromeda-optimised campaigns.