Andromeda - How it Works in 2026

Most perfume brands don’t fail because their ads are bad. They fail because they’re optimizing for signals that no longer matter. I know that sounds wrong at first, but it’s exactly what’s happening under the Andromeda update.

Reasoning

I ran a campaign recently for a fragrance brand that looked perfect on paper. The creatives were clean, the branding was strong, and engagement metrics were healthy. But the account wouldn’t scale. It stayed stuck in this loop of decent CTR and inconsistent conversions. The shift happened when we stopped chasing surface-level performance and started aligning everything with how Meta now prioritizes conversion behavior. Within days, performance stabilized and spend scaled without forcing it.

Visual Vs Behavioral

The core issue is that perfume brands are still thinking visually when they should be thinking behaviorally. This is a category where the product can’t be experienced digitally, which means the ad has to carry far more psychological weight. Most brands respond to that by making their ads more aesthetic. Slower shots, better lighting, cleaner packaging. But none of that answers the only question that matters to the buyer, which is who they become when they wear it.

When creative starts from identity instead of product, everything changes. You’re no longer presenting a fragrance, you’re positioning a version of the customer. The ads that convert aren’t the ones that look the best, they’re the ones that make the buyer feel understood. When someone sees an ad and thinks, “that’s me,” the click becomes inevitable. That’s what feeds the algorithm the signal it’s now prioritizing. Not curiosity, but intent.

Offer

This shift directly impacts how you structure your offer. If your entire funnel is built around selling a single bottle, you’re limiting your ability to scale before you even start. Customer acquisition costs in this category fluctuate, and if your average order value doesn’t give you room to absorb that, you’re forced to operate too tightly. The brands that scale are the ones that engineer value into the purchase. Discovery kits reduce friction. Bundles increase perceived payoff. Subscriptions extend lifetime value. These aren’t add-ons, they’re part of the acquisition system.

Targeting

Targeting has also changed in a way most brands are still resisting. The instinct is to narrow down audiences to improve efficiency, but the platform has moved in the opposite direction. Broad targeting works now, but only if your creative is doing its job. The algorithm doesn’t need help finding people anymore. It needs clear signals about who is converting and why. If your messaging is vague, no amount of targeting will fix it. If your messaging is sharp, broad audiences become an advantage instead of a risk.

Testing

Testing is where most of the damage happens. Perfume brands tend to be overly precious with their creatives. They launch a few polished ads, wait too long, and then make decisions based on incomplete data. That approach collapses under Andromeda. What works now is controlled volume. You need multiple variations entering the system consistently, each designed to test a specific angle or emotional trigger. The goal isn’t to guess the winner, it’s to create enough structured input that the system can find it for you.

Scaling then becomes a function of stability, not spikes. A lot of brands still chase short-term ROAS jumps, but those are unreliable signals. What matters now is whether an ad can hold a consistent cost per acquisition over time. When that happens, the system trusts the ad more and distribution increases naturally. If performance is volatile, scaling it only amplifies the instability. This is why some accounts feel like they hit a ceiling. It’s not a budget issue, it’s a consistency issue.

Conclusion

When you put all of this together, the path forward becomes clearer. You launch with multiple identity-driven creatives, not product showcases. You send traffic to an offer that increases order value and lowers friction at the same time. You let broad targeting do its job while focusing your effort on improving the quality of your creative signals. You test continuously, not cautiously, and you scale based on stable economics, not emotional reactions to performance swings.

At that point, Meta stops feeling unpredictable. The system starts behaving in a way that makes sense because you’re feeding it what it’s actually designed to optimize for. And once that alignment is in place, customer acquisition becomes far more controlled. Not easy, but predictable.