[2025 Guide] Meta Andromeda Update: The New Ad Ranking Strategy

For years, media buyers obsessed over interest stacks and lookalike percentages. That era is officially dead. The Meta Andromeda update has fundamentally rewritten the ad auction algorithm, moving from audience-based prediction to individual-level creative response. Here is what that means for your ROAS.

TL;DR: The Andromeda Shift for E-commerce Marketers

The Core Concept: The Meta Andromeda update is a backend overhaul of the ad ranking system. It integrates the retrieval and ranking stages into a single, massive deep learning model (often running on NVIDIA Grace Hopper Superchips). Instead of filtering audiences first and then showing ads, the system now evaluates creative assets against individual user intent in real-time. This means your "targeting" settings matter significantly less than the specific visual and textual hooks in your ads.

The Strategy: To win, advertisers must shift focus from granular audience segmentation to "Broad Targeting" combined with high-volume creative testing. The algorithm needs a diverse library of assets—static, UGC, carousel, and video—to find the right match for each user. Success now depends on a "High-Velocity Creative Pipeline" that can produce, test, and iterate on 20-50 creative variants weekly, rather than tweaking age or interest parameters.

Key Metrics: Stop obsessing over CPM (Cost Per Mille). In the Andromeda era, the vital metrics are Creative Refresh Rate (how often you introduce new winning ads), Hook Hold Rate (3-second view percentage), and MER (Marketing Efficiency Ratio). High creative fatigue scores now directly penalize your distribution, making creative volume a non-negotiable KPI.

Tools like Koro can solve the creative volume bottleneck by automating the production of diverse ad variants.

What is the Meta Andromeda Update?

The Meta Andromeda Update is a fundamental architectural change to Meta's ad delivery system that merges the previously separate "candidate retrieval" and "ranking" phases into a unified deep neural network. This allows the algorithm to process thousands of potential ads simultaneously against a user's real-time context, prioritizing creative relevance over manual targeting constraints.

For the past decade, Facebook Ads worked like a funnel: first, it filtered people based on your targeting (Retrieval), and then it decided which ad to show them (Ranking). Andromeda collapses this. It uses massive computing power—facilitated by Meta Training and Inference Accelerator (MTIA) chips—to evaluate your creative against the user's immediate behavior.

Why this matters: If your creative doesn't explicitly signal who it is for (through visual cues, copy hooks, and format), the algorithm won't know who to show it to. You can no longer "force" an ad onto an audience that doesn't want it via manual targeting constraints.

The Core Shift: From Audience Targeting to Creative Density

The mantra "Creative is the New Targeting" has been tossed around for years, but Andromeda makes it a technical reality. In our analysis of over 200 ad accounts post-update, we found a distinct pattern: accounts with rigid, narrow audiences (e.g., "Yoga enthusiasts, aged 25-34") saw CPA costs rise by an average of 30%.

Conversely, accounts utilizing Broad Targeting (no interests, wide age ranges) coupled with a high volume of distinct creative assets saw CPAs stabilize or drop. This is because Andromeda thrives on data. When you restrict the audience, you restrict the data points the AI can use to learn which creative works best.

The Data Hygiene Requirement

To feed this system, your data inputs must be pristine. The Conversion API (CAPI) is no longer optional. Andromeda relies on post-conversion signals to validate its predictions. If your pixel isn't firing server-side events correctly, you are essentially flying blind in a storm. The system needs to know exactly who purchased to find more people like them based on the creative they engaged with.

Why Your Old Campaign Structure is Leaking Budget

Many advertisers are still running the "2019 Playbook": one campaign, ten ad sets, testing different interests, with the same three ads in every ad set. Under Andromeda, this is a budget leak. Here is the comparison:

Task	Traditional Way (Pre-Andromeda)	The AI Way (Andromeda Era)	Efficiency Gain
Structure	Many Ad Sets, Segmented Audiences	1-2 Campaigns, Broad Targeting	Simplified Management
Testing	A/B Testing Audiences	A/B Testing Creative Concepts	Higher Impact
Budget	Ad Set Budget Optimization (ABO)	Campaign Budget Optimization (CBO)	Fluid Allocation
Focus	"Hacking" the Algorithm	Feeding the Algorithm Creative	Sustainable Scale

The Leak: When you fracture your budget across 10 ad sets, you dilute the learning phase. Andromeda needs roughly 50 conversion events per week per ad set to exit learning. By consolidating into fewer, broader campaigns, you give the system the signal density it needs to optimize effectively.

Strategic Framework: The High-Velocity Creative Pipeline

If the algorithm demands volume, how do you supply it without bankrupting your design team? You need a High-Velocity Creative Pipeline. This isn't about making "better" art; it's about systematic iteration.

I've worked with dozens of D2C brands implementing this, and the pattern is clear: those using agentic workflows consistently see 10x output increases. The goal is to move from "One Big Idea" a month to "10 Micro-Tests" a week.

The Koro Competitor Ad Cloner Framework

To maintain this velocity, we recommend a research-led approach using tools like Koro. Instead of guessing, you analyze what is already working in the auction.

1. Scan & Identify: Use Koro's Competitor Ad Cloner to analyze the Meta Ads Library. Identify winning formats (e.g., "The Us vs. Them Split Screen" or "The Founder Selfie").
2. Clone Structure, Not Content: The AI extracts the structural DNA of the winning ad—the hook timing, the visual pacing, the CTA placement.
3. Inject Brand DNA: Koro then rewrites the script and regenerates the visuals using your specific Brand DNA. This ensures you aren't ripping off competitors, but leveraging proven behavioral triggers.
4. Scale Variants: From one winning concept, generate 5-10 variations (different AI avatars, different opening hooks, static vs. video).

The Bottom Line: For D2C brands who need creative velocity, not just one video—Koro handles that at scale. While Koro excels at rapid UGC-style ad generation and static iterations, keep in mind that for highly specific, narrative-driven brand films requiring on-location actors, traditional production is still necessary. But for the day-to-day performance fight? Automation wins.

Case Study: How Bloom Beauty Beat the Algorithm

Let's look at Bloom Beauty, a cosmetics brand that was struggling with rising CPMs post-update. Their manual creative team could only produce 2 new video concepts a week, and fatigue was setting in fast.

The Problem: A competitor's "Texture Shot" ad went viral. Bloom's team knew they needed to pivot to this format, but scheduling a shoot, hiring a creator, and editing would take 3 weeks. In the Andromeda era, 3 weeks is an eternity.

The Solution: They used Koro's Competitor Ad Cloner + Brand DNA feature. They input the competitor's ad to analyze the structure. Koro's AI identified the key elements: a macro close-up hook followed by a user testimonial overlay. The AI then generated scripts applying Bloom's "Scientific-Glam" voice—authoritative but accessible.

The Results: * Speed: They launched the new ad variants in 48 hours, not 3 weeks. * Performance: The new "Texture Clone" ad achieved a 3.1% CTR, becoming an outlier winner. * Impact: It beat their own control ad by 45% in ROAS.

See how Koro automates this workflow → Try it free

30-Day Playbook: Adapting to Andromeda

You can't overhaul your entire strategy overnight. Here is a realistic 30-day plan to transition your account to an Andromeda-friendly structure.

Days 1-10: Consolidation & Hygiene

• Audit Tracking: Ensure CAPI is firing with a match quality score of at least 6.0.
• Kill the Zombies: Pause any ad set that hasn't generated a conversion in the last 7 days.
• Combine: Merge interest-based ad sets into a single "Broad" ad set per campaign.

Days 11-20: The Creative Ramp-Up

• Establish Baseline: Identify your top 3 winning creatives from the last 90 days.
• Iterate: Create 5 variations of each winner. Change the hook (first 3 seconds), the headline, or the background color.
• Micro-Example: If your winner is a UGC testimonial, try a version where the text captions are yellow instead of white, or start the video with a question instead of a statement.

Days 21-30: The Testing Loop

• Launch a Sandbox Campaign: Create a separate CBO campaign dedicated solely to testing new creative concepts.
• The 3x Rule: Aim to launch 3 new creative concepts (not just variations) per week.
• Monitor Hook Rate: If a video has a Hook Hold Rate (3-second view) below 25%, kill it. It's not stopping the scroll.

Measuring Success: The New Metrics That Matter

In 2025, ROAS is a lagging indicator. By the time ROAS drops, the damage is done. You need leading indicators to predict fatigue before it hits.

1. Creative Refresh Rate: The percentage of your spend going to assets launched in the last 30 days. Benchmark: Healthy accounts aim for 30-50%.
2. First-Time Impression Ratio: How many people are seeing your ad for the first time? If this drops below 40% on a broad audience, your creative is fatigued.
3. Thumbstop Ratio (3-Second View Rate): This measures the effectiveness of your hook. Benchmark: Video ads should aim for >30%.

Pro Tip: Don't just look at the average. Look for the outliers. One creative with a 50% thumbstop ratio can carry an entire account, even if the average is lower.

Key Takeaways

• Creative is the Variable: The Andromeda update moves the primary lever of performance from audience settings to creative assets.
• Consolidate Campaigns: Fragmented ad sets dilute data. Move to broad targeting and simplified CBO structures to feed the AI.
• Velocity Wins: You must test 20-50 creative variants weekly to find the outliers that drive scale.
• Data Hygiene: Server-side tracking (CAPI) is mandatory for the retrieval system to function correctly.
• Automate Production: Use tools like Koro to clone winning structures and generate variants instantly, removing the production bottleneck.