Two advertisers. Same target audience. Same budget. Same campaign structure. Their CPLs differ by 60%. That gap is creative quality — specifically, how Andromeda scores their ads before either one enters the auction. An ad rated low quality pays a premium to reach the same people. An ad rated high quality gets priority placement, lower CPMs, and preferential access to the users the system predicts will convert. If your creative testing process is 'make three versions of an ad and see what happens,' you're not running a testing framework. You're guessing at a system that already knows the answer. This piece covers what Andromeda actually scores, which formats earn high scores in the current environment, and how to build a testing cycle that systematically improves your creative quality signal over time.
The Four Dimensions of Andromeda's Creative Score
Andromeda evaluates creative across four dimensions that feed its internal quality scoring. Dimension 1 is Hook Strength — the degree to which the first 3 seconds of a video (or the primary visual of a static ad) captures attention and communicates a clear, relevant value proposition. Andromeda measures hook strength through watch-time completion rates and scroll-stop signals from its training data. Dimension 2 is Relevance Alignment — how consistently your ad's message, visual style, and value proposition align with the content of the destination landing page. Andromeda crawls your landing page and compares its signals against the ad creative, penalising disconnects. Dimension 3 is Conversion Prediction — Andromeda's model predicts the probability that a given creative will generate a genuine conversion event based on historical performance data from similar creative formats in your vertical. New accounts with limited history are benchmarked against vertical averages. Dimension 4 is Experience Quality — user experience signals from people who've previously interacted with your ads: dwell time on landing pages, bounce rates, form completion rates, and negative feedback rates (ad hidden or 'not interested' signals). High negative feedback is a Andromeda penalty that inflates CPMs for all future ads in the account.
- Hook strength in first 3 seconds: the most heavily weighted dimension for video creative in Andromeda's scoring
- Landing page relevance alignment: Andromeda crawls your LP and penalises creative-to-page message mismatches
- Conversion prediction score: Andromeda benchmarks creative against vertical conversion data to predict performance
- User experience signals: negative feedback (ad hidden, not interested) is a direct Andromeda CPM penalty trigger
- Static ad visual hierarchy: clear product representation, minimal text overlay, and strong contrast all contribute to quality score
The Creative Formats Andromeda Favours in 2026
Not all creative formats are equal under Andromeda's scoring architecture. Based on performance data from Meta's own A/B testing platform and independent agency observations across 2025–2026 campaigns, three creative formats consistently earn the highest Andromeda quality scores. First, native-style vertical video (9:16 ratio, 15–30 seconds, phone-quality production): Andromeda's training data reflects that Meta users engage more authentically with content that feels native to the feed rather than professionally produced. A founder talking directly to camera about their product's key benefit — filmed on a phone — routinely outperforms polished studio productions with 10× the production budget. Second, testimonial creatives with pre-qualification hooks: video or static ads that open with a specific customer result ('How we reduced our CPL from $120 to $45 in 60 days') followed by social proof content earn high conversion prediction scores because they pre-qualify viewers and historically convert at above-average rates. Third, problem-interrupt creatives that open by naming a specific, painful problem the target audience experiences, then position the advertiser's solution as the direct remedy. These earn strong hook scores because they trigger immediate pattern-interrupt responses in the target segment, generating above-average watch completion rates.
- Native vertical video (9:16, phone quality, 15–30s) consistently earns higher Andromeda scores than polished production in most verticals
- Testimonial creatives with specific result claims earn high conversion prediction scores from Andromeda
- Problem-interrupt hooks that name a specific audience pain trigger high scroll-stop rates that Andromeda rewards
- Carousel ads remain effective for product showcases but perform lower than video for lead generation objectives under Andromeda
- Text-overlay heavy static images receive Andromeda quality penalties — keep text under 20% of image area
Building a Creative Testing Framework for Andromeda
The most effective approach to Andromeda creative optimisation is a structured, hypothesis-driven testing framework rather than random creative rotation. The three-tier framework that works best: Tier 1 is Angle Testing — identifying which messaging angle (problem-led, benefit-led, social proof-led, authority-led) resonates with your target audience by running 3–4 distinct angle variants in a single broad ad set with CBO. Run for 7–10 days with a minimum $30/day budget, let Andromeda's algorithm optimise, and identify the winning angle based on CPL and conversion rate. Tier 2 is Format Testing — taking the winning angle and testing it across 2–3 formats (vertical video, static image, carousel) to identify which format Andromeda rewards most for your specific offer in your vertical. Tier 3 is Hook Testing — taking the winning angle and format combination and testing 3–5 distinct opening hooks to maximise Andromeda's hook score. Changes at the hook level have the largest CPL impact relative to production cost because hooks determine whether Andromeda's watch-time signals trigger a positive quality score update. The entire framework cycles on a 30-day basis, continuously feeding Andromeda higher-quality creative signals while systematically lowering CPL.
- Tier 1: test 3–4 messaging angles first — find what resonates before investing in production variety
- Tier 2: test winning angle across 2–3 formats to identify the format Andromeda rewards in your vertical
- Tier 3: test 3–5 hooks for the winning angle-format combination — hooks have the highest CPL impact per production dollar
- 30-day creative cycle prevents creative fatigue and provides Andromeda continuous fresh quality signals
- Kill underperformers after 7 days at $30/day minimum — insufficient data at lower budgets produces unreliable Andromeda signals
This is actually good news if you're not a massive advertiser. Andromeda's quality scoring rewards creative intelligence over raw budget. An SMB with a tight creative testing process and high Event Match Quality will consistently outperform a competitor spending 3× as much on a campaign that hasn't been updated in four months. The framework — angle testing, then format testing, then hook testing, cycling every 30 days — isn't complicated. It just requires discipline. Run it consistently and Andromeda's quality signals improve over time, lowering your CPMs without increasing spend.
Frequently Asked Questions
How does Andromeda score my ad creative before the auction?
Andromeda evaluates creative across four dimensions: hook strength (first 3 seconds for video, primary visual for static), relevance alignment with the destination landing page, conversion prediction based on historical data from similar creatives, and user experience signals from previous interactions with your ads. Higher scores earn lower CPMs in the auction — meaning creative quality directly affects your advertising costs, not just your click-through rate.
What is the most important creative variable to optimise for Andromeda?
For video ads, hook strength in the first 3 seconds is the highest-weighted Andromeda scoring variable. The opening hook determines Andromeda's watch-time and scroll-stop quality signals, which directly feed the creative quality score. For static ads, visual clarity and landing page alignment are the primary variables. In both cases, message-to-landing-page consistency is consistently underestimated by advertisers but heavily weighted by Andromeda.
How many creative variations should I run per ad set under Andromeda?
3–5 creative variations per ad set is the optimal range for Andromeda. Below 3 variations, the algorithm has insufficient creative diversity to run meaningful internal comparisons. Above 5–6 variations, budget gets spread too thinly across creatives for Andromeda to reach statistical confidence within a reasonable timeframe. Dynamic Creative Optimisation (DCO) can be used to test more variations simultaneously but works best when headline, image, and copy variations are small incremental changes rather than completely different creative concepts.