If you feel like your Meta Ads playbook from last year is broken, you’re not wrong. It is.
Since late 2025, performance has been… different. What worked before is suddenly sputtering. The culprit is a massive algorithm overhaul Meta calls ‘Andromeda,’ and it has completely rewritten the rules of the game. It’s no longer about finding the perfect audience; it’s about feeding a ravenous AI the right inputs.
I’ve spent the last six months re-tooling my entire approach across dozens of client accounts. Forget the theory. This is what’s actually driving results on the ground today.
Andromeda Is Here: Why Your Old Targeting Doesn’t Work
The biggest shift is that Meta has essentially taken the keys away from us when it comes to manual targeting. Andromeda isn’t just an update; it’s a replacement. Powered by NVIDIA’s ridiculously fast GH200 chips, it can process 10,000x more ad variations in parallel than the old system. It’s not trying to match your ad to an ‘interest’ you picked; it’s predicting conversion likelihood with a reported 87% accuracy.
How? By analyzing three layers of data: historical performance, real-time on-platform signals (likes, shares, saves), and most importantly, cross-account insights. Your days of stacking 20 interest layers are over. Broad targeting isn’t just an option anymore; it’s the entire strategy.
This makes your first-party data non-negotiable. The iOS ATT impact has already hit hard (65% signal loss), and with third-party cookies gone, the algorithm is starved for data. Setting up the Conversion API (CAPI) alongside your Pixel isn’t a ‘nice-to-have.’ I’ve seen it firsthand across my accounts: clients with a clean CAPI + Pixel setup see up to 24% more attributed conversions. That data feeds the algorithm, which in turn fuels the predictive expansion that makes broad targeting work.
Simplify Your Campaigns or Get Punished
My campaign structures from 18 months ago look like a crime scene today. The old method of creating dozens of ad sets to test micro-audiences now just starves the algorithm of the data it needs.
Today, my default structure for a lead gen or e-commerce brand is brutally simple:
- 1x CBO Sales Campaign: This is your workhorse. Use a broad, open Advantage+ audience. Seed it with your customer data, but let the algorithm expand. Don’t fence it in.
- 1x Awareness Campaign: Usually for video reach or ThruPlay. Build top-of-funnel consideration and feed the machine with engaged viewers.
- 1x Remarketing Campaign: A simple campaign targeting high-intent actions like View Content or Add to Cart. Again, Advantage+ audiences often outperform manual lists here.
That’s it. It feels wrong, I know. It feels like you’re not doing enough. But your job has changed. The single biggest mistake I see people make is tweaking things too early. The Andromeda system needs about 50 conversions per ad set within a week to exit the learning phase. Data from guys like Jon Loomer shows that a staggering 68% of advertisers make premature changes, constantly resetting the learning process and kneecapping their own results.
Be patient. Let it run for at least 7 days before you even think about touching it. If you’re running e-commerce, I’ve outlined the nitty-gritty of getting the most out of this new reality in a separate post on Meta Advantage+ Shopping: What Actually Works in 2026 (And What Doesn’t).
Creative Is the New Targeting
Let me be blunt: if you are still just testing different headlines and button colors on the same static image, you are actively losing money. Andromeda is designed to reward creative diversity and it actively penalizes similarity by jacking up your CPMs.
I saw one client’s CPL double in a week before we realized our ‘variations’ were just minor text tweaks on the same ad creative. The algorithm flagged it as fatigue.
The solution is radical creative diversity. You need to test completely different archetypes:
- A polished, cinematic founder story.
- A low-fi, phone-shot UGC testimonial.
- A trending meme format adapted to your brand.
- A clean motion graphics ad explaining your offer.
We now operate on a strict 14-21 day creative refresh cycle. The data doesn’t lie: once our ad frequency climbs above 3.5, we consistently see our cost-per-result increase by about 19%. This isn’t about avoiding boredom; it’s a direct lever on profitability.
And make it vertical. Over 90% of inventory is now 9:16 (Stories, Reels, Feed). Use Meta’s Flexible Ad Format to let the system automatically mix and match your assets for different placements. It lowers CPMs and saves you a ton of manual work.
The bottom line is this: the job of a media buyer has changed. We’re no longer just audience pickers and bid managers. We are data plumbers and creative strategists, responsible for feeding a powerful AI the fuel it needs to perform.
Want this applied to your campaigns? Book a free strategy call.