If your Meta ads have felt different this year — winners scaling faster, losers dying quicker, audience targeting mattering less than it used to — you're not imagining it. The engine underneath your account got replaced.
Meta's Andromeda algorithm is the biggest shift to ad delivery in years. It rebuilt the retrieval system that decides which ad to show which person, and it did it with roughly 10,000x more model complexity than the system it replaced. For ecommerce brands, that's not a back-end footnote. It changes what you build, how much of it you build, and where your edge actually comes from.
Here's the hard truth: the playbook most brands are still running was written for an algorithm that no longer exists. This is the new one.
What the Andromeda Algorithm Actually Is
Andromeda is the machine-learning system Meta uses to retrieve and rank the best ad for every impression. Think of it as the layer that sits between your campaigns and the user — it scans the entire pool of eligible ads and predicts, in milliseconds, which one is most likely to drive the outcome you're paying for.
The old retrieval system was relatively simple. It leaned heavily on audience signals: who you targeted, lookalikes, interest stacks, exclusions. You engineered performance by engineering the audience.
Andromeda flips that. With 10,000x the model complexity, it can read your creative with a level of nuance the old system never had. It understands the content of your ad — the hook, the format, the offer, the visual — and matches it to the person most likely to respond. The audience is now an output of the creative, not an input you control.
That single change is why everything downstream feels different.
Audience Targeting Is No Longer Your Lever
For a decade, Meta media buyers competed on targeting. Better lookalikes, smarter interest layering, tighter exclusions. That era is over.
Under Andromeda, broad targeting consistently matches or beats hand-built audiences, because the algorithm is now good enough to find your buyer on its own — if you give it the creative signals to do it. Lookalike audiences, the workhorse of ecommerce scaling for years, are effectively dead as a growth lever. They still exist in the interface. They just don't give you the edge they used to.
This is the part most operators get wrong. They respond to softer targeting performance by adding more targeting — more audiences, more exclusions, more manual control. That's fighting the algorithm. The brands winning under Andromeda do the opposite: they go broad, they let the system retrieve, and they pour their energy into the one input that still moves the needle.
That input is creative.
Creative Is the New Targeting
Here's the mental model shift: your creative is your targeting now. Every ad you launch is a signal that tells Andromeda who to find. A founder-story video signals a different audience than a UGC unboxing, which signals a different audience than a catalog ad showing 12 SKUs. The algorithm reads those differences and routes accordingly.
That means creative diversity isn't a nice-to-have. It's how you give the algorithm enough surface area to do its job. A campaign with three near-identical static ads gives Andromeda almost nothing to work with. A campaign with a dozen genuinely different concepts gives it room to find pockets of demand you'd never have targeted manually.
The brands that internalize this stop asking "who should we target?" and start asking "how many distinct angles can we put in front of the machine?"
The New Volume Standard: 8–12 Creatives Per Campaign
Under the old system, you could run a campaign on three or four ads and ride a winner for weeks. Andromeda burns through that. Because it distributes creative faster and more precisely, it exhausts a small ad set quicker — and it has nothing to fall back on when those ads fatigue.
The new floor for ecommerce is 8–12 distinct creatives per campaign. Not 12 minor variations of one ad — 12 different angles, formats, and hooks. That gives the algorithm the diversity it needs to keep finding fresh audience and keeps your campaign from collapsing the moment your top performer tires.
This is a production problem before it's a media problem. If your creative pipeline ships two or three new concepts a month, you can't feed an Andromeda-era account. The constraint on your growth has moved from your media buyer's targeting skill to your team's creative output. Plan your operation around that.
Catalog Ads Are Outperforming Static by 23%
One concrete shift worth acting on now: catalog ads (dynamic product ads pulling from your feed) are outperforming standalone static creative by roughly 23% under Andromeda.
It makes sense once you understand the retrieval logic. Catalog ads give the algorithm thousands of product-level combinations to test against each impression — which is exactly the kind of variety Andromeda is built to exploit. Instead of one static image, the system can match the specific product most likely to convert that specific person.
For ecommerce brands, the takeaway is direct: if your product feed is neglected, you're leaving performance on the table. A clean, complete, well-segmented catalog is now a core performance asset, not a Shopping-tab afterthought. Audit your feed. Fix missing fields, bad titles, and stale inventory. Then make catalog ads a permanent pillar of your account structure alongside your concept-driven creative.
The 2026 Andromeda Playbook for Ecommerce
Pulling it together, here's what running a Meta account under Andromeda actually looks like:
→ Go broad on targeting. Stop hand-building audiences as your primary lever. Let the algorithm retrieve. Use exclusions sparingly and only where they protect margin (e.g. recent purchasers).
→ Ship 8–12 distinct creatives per campaign. Different angles, not different thumbnails. Treat creative volume as the real budget.
→ Build a real catalog ads pillar. Clean the feed, segment it, and run dynamic product ads as a permanent part of the account — they're beating static by ~23%.
→ Diversify formats deliberately. Video, UGC, static, founder story, catalog. Each format is a different targeting signal. Variety feeds retrieval.
→ Refresh on the algorithm's clock, not yours. Andromeda exhausts creative faster than the old system. Build a rotation cadence that assumes shorter lifespans, not the 4–6 week cycles you learned in 2023.
→ Move spend on signal, not sentiment. The system surfaces winners fast. Let it. Kill losers fast, scale winners fast, and keep the concept pipeline full so you always have the next test ready.
The Real Shift Isn't Technical — It's Operational
It's tempting to treat Andromeda as a media-buying update. It isn't. The accounts that win in 2026 aren't the ones with the cleverest targeting — that lever is gone. They're the ones with a creative engine that can reliably produce volume and variety, paired with a clean catalog and a media buyer who knows how to get out of the algorithm's way.
That's the uncomfortable part for a lot of brands. The bottleneck moved. It used to live in the ad account. Now it lives in your content operation. The Andromeda algorithm rewards the brand that can ship — and quietly starves the one that can't.
Get your creative output and your product feed right, and Andromeda becomes the best media buyer you've ever had. Ignore them, and no amount of in-platform tweaking will save your ROAS.
Ready to Scale Your Ecommerce Ads?
If you’re spending $10K+ on Meta ads and want an agency that actually knows ecommerce — book a free strategy call. We’ll audit your account and show you where the growth is.
