How The Baer Edge Prepared For The Biggest Shift in Facebook Advertising History — and Why Our Clients Are Already Winning in 2026

The Algorithm Changed. We Were Ready.

In late 2024, Meta quietly began deploying a new ad delivery system called Andromeda. By October 2025, the global rollout was complete — and Facebook and Instagram advertising had fundamentally and permanently changed.

Andromeda didn’t tweak the old system. It replaced it entirely. Meta’s engineering team built a new personalized ads retrieval engine powered by deep neural networks, the NVIDIA Grace Hopper Superchip, and Meta’s own Training and Inference Accelerator hardware — creating a system with 10,000 times the model complexity of the previous delivery engine. The system now scans tens of millions of candidate ads in milliseconds, analyzing hundreds of behavioral signals for every single impression opportunity, and matches ads to individual users based on real-time engagement patterns, watch time, purchase history, save behaviors, and content interactions.

The shift was seismic. For years, Facebook advertising success depended on precise audience targeting — carefully selecting demographics, interests, behaviors, and lookalike audiences to control who saw your ads. Andromeda flipped that equation. Under the new system, Meta’s AI controls the targeting. The algorithm evaluates your creative first — its visual composition, messaging, format, narrative tone, and engagement signals — and then determines which individual users are most likely to respond to it.

Meta confirmed this directly in their April 2025 blog: the focus has shifted from niche targeting to creative diversification as the primary lever for finding the most relevant audiences.

Creative is no longer part of the strategy. Creative is the strategy.

At The Baer Edge, we didn’t read about this shift and scramble to react. We saw it coming, built for it throughout 2025, and entered 2026 with a creative production infrastructure specifically engineered for the way Facebook and Instagram advertising works now. Here’s the full story of how we did it — month by month — and why it matters for every business running ads on Meta’s platforms.

January – February 2025: We Saw the Signals Before the Industry Had a Name for Them

While most agencies were carrying their 2024 playbooks into the new year, we were already tracking anomalies across our client accounts that didn’t add up.

Starting in late 2024, campaigns that had been performing reliably for months began behaving erratically — inconsistent reach, unpredictable CPMs, ad sets that would surge one week and flatline the next. Nothing had changed on our end. Nothing had changed on our clients’ end. Something was shifting underneath the platform itself.

We pulled performance data across every account we managed and started looking for patterns. What we found was the earliest indicator of what was coming. Campaigns with limited creative variety — three to five similar ad variations running against carefully targeted audiences — were declining fastest. Campaigns where we had already been testing multiple creative angles, visual styles, formats, and messaging hooks were holding steady or improving.

By February, early industry reporting confirmed what our data was telling us. Meta was rolling out a new AI-driven ad retrieval engine that evaluated creative quality and diversity as a primary ranking signal before ads even entered the auction. This wasn’t a feature update or a minor algorithm adjustment. It was a fundamental re-architecture of how every ad on Facebook and Instagram gets selected, delivered, and shown to users.

We understood immediately what this meant for our clients — and for the industry. The old model of building the perfect audience and serving it a handful of polished ads was dying. The new model demanded something entirely different: a continuous pipeline of meaningfully diverse creative assets that gave the algorithm enough variation to learn, optimize, and deliver at the individual user level.

In February 2025, we made a strategic decision that would define our entire year and reshape how we serve our clients. We committed to building a new creative production infrastructure from the ground up — not as a reaction to Andromeda, but as a proactive investment in the advertising model we believed was coming for everyone.

March – April 2025: Building The Creative Engine

By March, we had confirmed through direct testing across multiple client verticals what the industry was only beginning to understand. The old Facebook advertising model — create a few strong ads, test them against carefully segmented audiences, scale the winners — was producing diminishing returns that would only accelerate as Meta’s rollout continued.

The new model demanded volume with genuine variety. And not the kind of “variety” most agencies were used to — swapping a headline, changing a background color, adjusting a CTA button. Meta’s new system is far more sophisticated than that. Andromeda doesn’t just read your ad copy. It pattern-matches visual cues, analyzes video structure, evaluates narrative tone and frame dynamics, and measures the ratio between views and saves. If you upload 12 ads that look and feel similar — even if the words are different — Andromeda groups them together and treats them as a single creative option. If it determines the core concept of that cluster is weak, it suppresses all of them simultaneously.

This is why so many advertisers saw their “winning” ads suddenly die in early 2025. The algorithm had judged the entire creative cluster as fatigued, and there wasn’t enough meaningful diversity to offer it an alternative.

We built what we now call The Creative Engine — an in-house production and strategy system designed specifically for the demands of AI-driven advertising on Meta’s platforms. It’s built on four core principles.

First, volume with genuine variety. We produce a high number of creative assets per campaign cycle, but every asset is designed to represent a fundamentally different angle — different emotional triggers, different storytelling approaches, different visual styles, different talent, different formats, different calls to action. We’re not creating iterations. We’re creating distinct concepts that prevent the algorithm from clustering our ads together and limiting delivery.

Second, rapid iteration cycles. Under Andromeda, even top-performing ads fatigue faster than they did under the old system — typically within two to four weeks. Our Creative Engine is designed for biweekly to monthly creative refreshes, ensuring our clients’ campaigns always have fresh material entering the system before performance decay sets in. We don’t wait for an ad to die before replacing it. We have the next wave ready before the current wave peaks.

Third, data-informed creative strategy. Every creative variation we produce is informed by performance data from previous cycles. We track which hooks, formats, styles, and messages drive the strongest signals — not just clicks, but the deeper engagement metrics Andromeda rewards: watch time, saves, shares, meaningful comments, and conversion actions. Those insights feed directly back into the next round of production, creating a compounding intelligence loop that gets smarter with every cycle.

Fourth, format diversity by design. Andromeda doesn’t just decide who sees your ad — it decides where and in what format. A user scrolling Stories at 10 PM responds differently than the same user browsing their Feed at lunch. Our Creative Engine produces every campaign message across multiple formats — short-form video, Reels, carousel posts, static images, Stories content — giving the algorithm maximum flexibility to match the right format to the right user in the right context.

By April, The Creative Engine was operational across our entire client base. While the industry was still diagnosing why campaigns were underperforming, we were already producing and testing at the volume and variety Andromeda demanded.

May – June 2025: Proving It Works Across Every Vertical We Serve

May and June were our validation months. Meta’s Andromeda rollout was accelerating globally, and the performance gap between creative-first campaigns and legacy-structure campaigns was widening by the week.

We ran controlled comparisons across every client vertical we serve — VR entertainment venues, automotive dealerships, medical spas, professional services — measuring our Creative Engine approach against traditional campaign structures. The results confirmed everything we had built toward.

Campaigns running 12 or more meaningfully different creative variations consistently outperformed campaigns running three to five similar variations. Cost per lead dropped. Conversion rates improved. Ad delivery became more stable and predictable because the algorithm had enough creative diversity to continuously find the right message-to-person match across different audience segments.

We also confirmed something critical about account structure that many agencies were still resisting. Andromeda performs best with simplified campaign architectures. The old approach of running dozens of hyper-segmented ad sets — each targeting a narrow slice of audience with its own budget and creative — was not just outdated. It was actively hurting performance by fragmenting learning and preventing the system from optimizing effectively.

We consolidated our clients’ campaigns into fewer, broader campaigns with diverse creative. We adopted Advantage+ placements and Campaign Budget Optimization, allowing Meta’s automated systems to handle audience selection, placement optimization, and budget allocation. For most campaigns, we went broad on targeting — removing interest restrictions, age selections, and gender filters — and let the algorithm do what Andromeda was designed to do: find the highest-value prospects based on creative engagement signals rather than static audience definitions.

The exception — and this matters enormously for the local businesses we serve — is geographic targeting. A VR entertainment venue in San Diego doesn’t want leads from Phoenix. An automotive dealership in Fort Worth doesn’t need clicks from Chicago. For local businesses, we use geographic targeting as a boundary to restrict delivery to the service area, but go broad on everything else within that geography. This gives Andromeda enough data space to optimize while keeping results relevant to the business’s actual market.

Meta’s own testing data aligned with what we were seeing. Campaigns using Advantage+ with broad targeting showed up to 10% lower cost per lead compared to traditional manual campaigns. Our results matched this — and in several client accounts exceeded it, because The Creative Engine was supplying the creative volume and diversity the system needed to perform at its best.

By June, we had fully transitioned every client account to Andromeda-optimized structures. The creative-first approach wasn’t an experiment anymore. It was our operating model.

July – August 2025: Refining What the Algorithm Actually Rewards

With months of live performance data accumulating across every vertical we manage, July and August became our refinement phase. We had built The Creative Engine and proven it worked. Now we were learning — in granular detail — exactly what Andromeda rewards and what it punishes.

What we confirmed works: authentic, relatable creative generates stronger early engagement signals than polished, studio-produced advertising. This finding was consistent across every vertical — VR entertainment, automotive, medical spas, professional services. User-generated style content, real customer testimonials, founder-led and staff-led storytelling, and behind-the-scenes footage consistently outperformed high-production-value ads that looked and felt like traditional advertising.

This makes sense when you understand how Andromeda evaluates creative. The algorithm measures engagement depth — not just whether someone clicked, but whether they watched, how long they watched, whether they saved the content, whether they shared it, whether they commented meaningfully, and whether they took a conversion action. Content that feels authentic and human generates those deeper engagement signals more reliably than content that feels like an advertisement. People scroll past ads. They stop for stories.

We also confirmed the critical importance of format mixing within campaigns. Providing Reels, carousel posts, static images, and Stories content within the same campaign gave the algorithm dramatically more options for matching creative to placement and user context. Campaigns with format diversity delivered more stable results and more efficient spend than campaigns limited to a single format.

What we learned to eliminate: the practice of making minor variations and calling it “testing.” Under the old system, you could change a headline, swap a model, adjust a color scheme, and call that an A/B test. Under Andromeda, those kinds of surface-level changes don’t register as diversity. The system pattern-matches at a much deeper level — visual composition, narrative structure, emotional tone, creative concept. If your ads share the same fundamental concept, Andromeda treats them as one option regardless of the cosmetic differences.

We restructured our entire creative testing framework around what we call conceptual testing. Instead of testing variations of the same idea, we test fundamentally different ideas — different emotional appeals, different customer pain points, different storytelling structures, different visual worlds. Each asset targets a genuinely different psychological angle, ensuring the algorithm has real options to work with rather than a cluster of near-duplicates.

We also learned something about creative velocity that became a core operating principle. Under Andromeda, the best-performing ad in your account today has a shelf life. Two to four weeks, typically. Sometimes less. The agencies and businesses that treat creative production as a periodic event — a big photoshoot every quarter, a batch of ads every few months — are fighting a losing battle against fatigue in a system that cycles through creative faster than ever before. Our Creative Engine is designed for continuous production. New creative enters the system before existing creative peaks, ensuring our clients’ campaigns never hit the fatigue wall that causes the dramatic performance drops so many advertisers experienced in 2025.

September – October 2025: The Industry Catches Up. We’re Already Ahead.

By September, Meta’s Andromeda rollout was reaching completion globally. The industry was fully awake to the shift — agencies everywhere were publishing explainers, running emergency webinars, scrambling to retrain their teams, and trying to figure out how to produce the creative volume Andromeda demanded.

We were already months ahead.

Our clients’ campaigns had been running on Andromeda-optimized infrastructure since the spring. While competitors were experiencing the performance disruptions that come with catching up — the learning curves, the failed experiments, the trial-and-error of figuring out what works under a fundamentally new system — our accounts were optimized, producing, and performing.

The competitive advantage was measurable and visible in the data. Our clients’ campaigns were delivering more consistent results at stable or decreasing costs during a period when industry benchmarks showed widespread cost increases and performance volatility. The Creative Engine’s continuous refresh cycle meant our clients’ campaigns never hit the creative fatigue wall that was causing dramatic performance drops for advertisers still refreshing monthly or quarterly.

October brought two significant developments. First, Meta officially confirmed the global rollout completion of Andromeda and released additional details about how the system evaluates creative. Meta introduced new metrics — Creative Similarity scores and Top Creative Themes — that validated our entire approach. A high Creative Similarity score means your ads lack diversity, and the algorithm responds by raising your CPMs because it views the content as repetitive and fatiguing. We had been building against exactly this dynamic since March.

Second — and this was a development we had been anticipating — Meta began revealing details about companion AI systems working alongside Andromeda. The company disclosed that a system called GEM (Generative Ads Recommendation Model) had been running in the background since the second quarter of 2025. GEM is Meta’s most advanced advertising AI — a foundation model built at the same scale as large language models, trained on thousands of GPUs, analyzing both paid ad interactions and organic content engagement across Facebook, Instagram, Messenger, and other Meta properties.

The relationship between Andromeda and GEM matters for understanding where advertising is headed. Andromeda handles retrieval — deciding which ads are eligible to be shown. GEM is the foundation model that provides predictive intelligence across the entire system, learning what works and sharing those insights across all of Meta’s platforms and placements. Where previous systems optimized Instagram and Facebook separately, GEM creates unified intelligence. Meta reported that GEM delivered a 5% increase in ad conversions on Instagram and a 3% increase on Facebook Feed — with those improvements doubling in the third quarter as the model was refined.

For our clients, this development reinforced the value of everything we had built. GEM rewards exactly what The Creative Engine produces: diverse, high-quality creative that generates strong engagement signals across multiple formats and placements. The better the creative inputs, the more effectively GEM and Andromeda work together to find the right audience and deliver the right message.

November – December 2025: Building for What Comes Next

As 2025 closed, the advertising industry was still processing the implications of Andromeda. Most agencies were in reactive mode — publishing “what changed” guides, running “how to adapt” workshops, and beginning the long process of restructuring their creative production capabilities.

We spent November and December doing something different: preparing for the next wave.

Meta has made it clear that Andromeda and GEM are the foundation, not the destination. The company is developing increasingly sophisticated generative AI tools for ad creation. More than a million advertisers were already using Meta’s GenAI tools to create over 15 million ads per month by the end of 2025 — with Meta estimating a 7% increase in conversions for businesses using AI-generated images. The trajectory points toward a future where advertisers can provide a product URL, a budget, and basic creative direction, and Meta’s AI generates and optimizes the entire campaign automatically.

This development reinforces rather than threatens our approach — and this is a critical distinction that separates agencies built for the future from agencies built for the past.

AI can generate creative at scale. What it cannot replace is the strategic thinking that determines what that creative should say, who it should speak to, why it matters, and how it aligns with a brand’s identity and business objectives. The agencies that will thrive as Meta’s AI capabilities expand are the ones that understand creative strategy at a deep level — the ones that know which messages resonate with which audiences, which emotional triggers drive action in specific verticals, and which brand narratives build long-term customer value versus short-term clicks.

The Creative Engine isn’t just a production system. It’s a strategic framework. Every creative asset we produce is informed by data, aligned with brand strategy, grounded in our knowledge of each client’s specific market and customer psychology, and designed to perform within an AI-driven delivery system. That foundation becomes more valuable, not less, as Meta’s AI capabilities expand.

We also deepened our organic content capabilities through November and December. Andromeda’s changes extend well beyond paid advertising. Meta’s updated ranking systems now reward authentic, engagement-driven organic content with greater reach and discovery. Content that generates saves, shares, meaningful comments, and extended interaction gets amplified by the algorithm — and those organic engagement signals reinforce paid campaign performance within the same ecosystem.

Some publishers reported referral traffic increases of up to four times compared to the previous year. For businesses, this means organic content strategy and paid advertising strategy are no longer separate disciplines. They’re two halves of the same system. Our Creative Engine now produces integrated strategies — paid creative assets and organic content designed to work together, creating a compounding effect where strong organic performance feeds paid campaign efficiency and vice versa.

January 2026: Where We Stand — and Where Our Clients Stand

As we enter 2026, the advertising landscape on Facebook and Instagram is defined by a single reality: this is now a creative-first discipline. The old levers — audience segmentation, manual bid adjustments, interest-based targeting, complex account structures — have been replaced by a new imperative: produce diverse, high-quality, authentic creative at a volume and pace that feeds Meta’s AI systems and keeps campaigns performing consistently.

This is exactly what we spent all of 2025 building.

Our clients enter 2026 with The Creative Engine fully operational — producing the creative diversity that Andromeda demands at the velocity the system requires. Their campaign structures are optimized for the new architecture — simplified, broad, algorithm-friendly setups that let Meta’s AI do what it does best. Their creative refresh cycles prevent fatigue before it impacts performance. Their conversion tracking infrastructure feeds clean, accurate data back to the algorithm, ensuring the system learns and improves with every campaign cycle. And their organic and paid strategies are integrated to maximize visibility in an ecosystem that rewards authentic engagement across both channels.

Here’s what we see coming for the rest of 2026 — and what we’re actively preparing our clients for.

Creative volume and velocity will become even more important as competition for Andromeda’s retrieval attention increases. The businesses producing 15 to 20 or more meaningfully different creative variations per campaign will consistently outperform those still running five or six similar ads. The gap will widen, not narrow.

Authentic, human content will continue to outperform polished advertising. Andromeda and GEM measure engagement depth, and people engage more deeply with content that feels real. Brands that invest in genuine storytelling — real customers, real staff, real experiences, real behind-the-scenes moments — will generate the engagement signals that the algorithm rewards most.

Video — especially short-form — will dominate creative performance. Andromeda’s signal analysis is most effective with video content, where it can measure watch time, completion rates, replays, and interaction patterns. Reels and short-form video will be the highest-performing formats for most advertisers throughout 2026.

The integration of AI creative tools with human strategic direction will define the most effective agencies. AI accelerates production. Human expertise provides the brand voice, emotional intelligence, market knowledge, and strategic positioning that determine whether that production actually drives business results. The agencies using both — AI for speed and scale, humans for strategy and soul — will deliver the strongest outcomes.

Organic content strategy will become inseparable from paid advertising strategy. Andromeda’s ecosystem increasingly rewards brands with strong engagement signals across both channels. Businesses that treat organic as an afterthought while pouring budget into paid ads will underperform businesses that build integrated strategies where organic and paid reinforce each other.

And Meta will continue reducing manual controls. Detailed targeting options have been deprecated throughout 2025. Advantage+ campaigns where Meta’s AI handles most decisions are being pushed as the default. This trend will accelerate. The advertisers who are comfortable letting go of granular manual control and instead focusing on creative quality, strategic direction, and data integrity will outperform those who try to hold onto the old way of doing things.