Your Snack Brand Is Targeting the Wrong Thing on Meta (And It's Killing Your CPA)

Most food brands running Meta ads are spending real time building interest audiences. "Healthy snacks." "Foodies." "Organic living." They layer in behavioral qualifiers, stack demographics on top, and feel good about it because it feels controlled. Methodical. Like they know something.

They don't. And neither did I, until I stopped doing it.

I know that sounds counterintuitive. Bear with me.

What Actually Runs Ads in 2026

About two years ago I was helping run ads for a client selling a productivity app. We were stacked on interests — "entrepreneurship," "self-improvement," all the obvious ones. Results were decent. Ceilings came fast.

On a whim, we stripped everything. Ran broad. Built one creative around a specific freelancer drowning in client work, unable to figure out where her time was going. Within a week, CPA dropped 31%. The buyers who came in were nurses, teachers, small restaurant owners. Nobody we would have targeted. The algorithm found them because the creative spoke to something they already felt.

That story has nothing to do with productivity apps. It has everything to do with how you should be running ads for your beef jerky, your protein chips, your hot sauce subscription.

The Algorithm Is Your Targeting Department Now

Meta's Andromeda system is processing billions of behavioral signals in real time. What people pause on. What they click. What they buy at midnight when they're hungry and bored and scrolling. These signals don't live inside interest categories. They're messy, contextual, and constantly shifting.

When you build a manual audience for your snack brand, you're telling a billion-signal model that you know better than it does who craves your product. You don't. I don't. Nobody does.

The creative is your targeting now. The person in your video, the hook in the first three seconds, the craving you trigger visually — those are the signals Andromeda reads. It identifies who watches all the way through, who slows down at the texture shot, who clicks after seeing the ingredient list. Then it finds more of them. That's the loop. And manual interest targeting breaks it.

What Food Creative Has to Do That Other Categories Don't

Food is physical. Emotional. Visceral. The purchase decision for a $14 bag of trail mix or a $45 snack subscription box isn't rational. It's triggered by a craving, by recognition, by the three-second moment when your ad makes someone's mouth water before their brain catches up.

That's your job. Not explaining your product. Triggering the craving first.

This means your creative decisions — texture shots, the sound of a crunch, a hand pulling apart your product in natural light — are not production choices. They are targeting decisions. They determine who leans in and who scrolls. They train Andromeda on who your buyer actually is.

If your video opens with a white-label talking-head explaining your sourcing story, you've already lost. If it opens with a close-up of melted dark chocolate hitting sea salt caramel, you've just told Meta's system exactly who to find more of.

The Mistake Food Brands Make in Week One

Here's the pattern I see constantly. Brand launches broad creative, gets mediocre results in week one, panics, layers in interests to "fix" things. CPA ticks down slightly because the audience is smaller. The ceiling drops. Scale disappears. They call it a strategy.

It's not. It's a ceiling you built yourself.

The other failure mode is building creative around what the brand is proud of instead of what the buyer actually feels. Your sourcing story matters to you. Your customer doesn't care yet. What they care about is: does this solve my 3pm slump? Can I eat this without guilt? Will my kids actually finish it?

Put their felt problem in the first three seconds. The product details come after you have their attention.

How to Build Creative That Acts as Targeting

Hook by craving, not by product. The first line of your ad should name a feeling, a moment, or a problem your buyer already has. "You hit 3pm and you're reaching for garbage again." "Every healthy snack your kids have tried ends up in the trash." That's not a claim about your product. That's a mirror held up to your customer. They lean in because they feel seen.

Show the product doing something. A bag sitting on a counter is not creative. Someone opening it, tasting it, reacting genuinely — that's a signal generator. Andromeda is watching engagement. Give it something to work with.

Cast your actual buyer. If you're selling high-protein snacks to women 35–50 managing their weight while feeding a family, put that woman in your video. People respond to faces that look like theirs. Stronger responses give the algorithm better data. Better data lowers your CPA.

Use your packaging as a prop, not a prop. Packaging that's visible throughout — on a counter, being carried, being stacked in a pantry — builds brand recall while giving Meta visual anchors to match against lookalike behavior. It works even when you're running broad.

What Your Testing Framework Should Look Like

Stop testing audiences. Test creative angles. Run three to five distinct hooks at the same time, each targeting a different craving or problem state. Keep the offer identical. Let performance tell you which angle resonates with the largest segment of buyers Andromeda can find.

In the first two weeks, you're not optimizing for ROAS. You're feeding the algorithm. Your job is to generate quality engagement events — clicks, adds to cart, purchases — so the learning phase exits clean. Cheap, broad creative that generates real clicks beats expensive, narrow creative that generates filtered ones. Every time.

Once you have a winning angle, you scale by expanding creative variation, not by widening audiences. More hooks. Different formats. Reels vs. static. UGC vs. polished. The angle stays. The container changes. That's how you find scale without torching your CPA.

The Unit Economics Your Creative Has to Support

Know your numbers before you launch anything. If your AOV is $28 and your gross margin is 45%, your target CPA shouldn't exceed $12–15 on a single purchase basis if you're not modeling LTV. If you have subscription data and your 90-day LTV is $85, you can afford to acquire at $30 and still print money.

The creative strategy isn't separate from the unit economics. It's what allows you to bid competitively in the auction and still be profitable. A brand that knows its LTV:CAC ratio can afford to test more angles, survive the learning phase, and outlast competitors who are scared of week-one CPAs.

Your snack brand isn't losing because Meta is expensive. It's losing because the creative isn't generating enough signal for Andromeda to work with, and you're compensating by narrowing the audience and wondering why nothing scales.

Close: What Predictable Customer Acquisition Actually Looks Like

Predictable acquisition for a food brand on Meta is not a complex targeting architecture. It's a creative engine that consistently produces ads with strong engagement signals, a broad delivery setup that lets Andromeda do its job, and a unit economics model that tells you exactly how much you can pay to win a customer and still grow.

The brands that figure this out stop competing for audiences and start competing on creative quality. They run more tests, kill losers faster, and scale winners harder. They know their CPA targets by product and by season. They know which visual hooks drive impulse buys and which ones drive subscriptions.

That's the system. Build creative that trains the algorithm. Let the algorithm find your buyer. Know your numbers well enough to scale without flinching.

Everything else is noise.