What Meta's Andromeda Update Did to Ad Targeting in 2026

Andromeda quietly killed the targeting playbook most agencies still run in 2026. Interest + lookalike + age range used to be the job. Now Meta barely reads those fields. The signal it actually weighs in 2026 lives in the first 2 seconds of your creative, not in your audience builder.

I’ve been running Meta ads since I was 12 years old. My first big break came at 15 when the Would You Rather app I built hit 5M+ downloads and #1 US App Store using Meta as the entire growth engine. I’ve watched every major algo shift since the early days of FB Ads Manager, before most of you reading this had a Facebook account.

Andromeda is the biggest one in years. Most agencies haven’t caught up. They’re still blaming the algorithm when their CPAs creep past $200 per booked call.

Here’s what actually changed and what to do about it.

The shift: creative IS the targeting signal now

Pre-Andromeda, your audience setup carried 60-70% of the targeting weight. You picked an interest cluster, Meta found people inside it, and your creative just had to convert them.

Post-Andromeda, that ratio flipped. Your creative now carries 60-70% of the targeting weight. Meta watches who reacts to your hook, scrolls past it, swipes through it, and taps it. Then it goes and finds more people who behave like the reactors.

The audience setup is now mostly a fence around the search space. The creative does the actual hunting inside that fence.

Here’s why this matters in dollars. When an agency owner pings me saying his $5K/mo Meta spend is producing $215 cost per booked call, his first instinct is always to fix the audience. He wants to add new lookalikes or tighten the interests or layer a custom audience on top.

That used to work. It doesn’t now.

I took Brandon, an agency owner with that exact CPA problem, from $215 per booked call down to $30 in 2 weeks. We didn’t touch his audience setup once. We rebuilt three creatives. The audience signal flowed through the creatives.

That’s the shift in plain English. The creative tells Meta who you want and Meta delivers them.

What agencies are still doing wrong

Walk into 8 out of 10 agency ad accounts right now and you’ll see the same anti-pattern: 5 lookalikes layered on top of 3 interest stacks, detailed targeting expansion turned on, and Advantage+ shopping campaigns running parallel to broad ABO setups.

That’s audience-side complexity. The creative side is usually too simple to balance it out.

The creatives are usually one polished UGC, one talking head, maybe a static. All three making roughly the same point with roughly the same hook.

This was a winning formula in 2022. It’s a budget incinerator in 2026.

When Meta’s targeting brain was dumb, you compensated with audience precision. Now Meta’s brain is sharp. It needs creative variety to feed on. Three lookalike-style ads all saying “we book qualified calls” gives Meta nothing to learn from.

The other thing agencies miss is automatic enhancements. Meta’s auto-enhance feature is on by default. It distorts your video, adds music over your voice, crops your frame, and adjusts your color. I’ve literally seen it add pop music over a founder’s testimonial that was supposed to be intimate. The algorithm then gets a confused signal because the version it’s testing isn’t the version you built.

Turn auto-enhancements off. CPA drops 30-40% on the same creative the next day in most accounts I’ve audited. That’s a 5-minute fix nobody’s clients have done.

The 3 creative levers Andromeda actually rewards

If creative is the targeting signal, the question becomes: what variables in the creative does Meta actually pay attention to?

I’ve stress-tested this across DTC, agency, and B2B SaaS accounts over the last 18 months. Three levers move the targeting signal more than anything else.

Lever 1: the first 3 seconds

The first 3 seconds is where the targeting decision happens. Meta watches who keeps watching past second 3 and uses that population as the seed for delivery.

Most agencies are still using the first 3 seconds as setup time. “Hey what’s up everyone, today I’m going to show you...” Three seconds of nothing. By the time the actual hook lands at second 5, Meta already classified the early viewers as not-interested and moved on to a different audience.

Your hook needs to hit at second 0. The viewer’s pattern interrupt has to happen before they’ve decided whether to scroll. We took a creative for an agency client whose hook was at second 6 and rebuilt it with the same content but with the hook moved to second 0. CPA dropped from $180 to $74 in 9 days on the same offer and audience and product.

Lever 2: visual specificity

Andromeda watches what you show, not just what you say. A creative that visually displays a CRM screen, a Calendly screenshot, an Ad Manager dashboard, a bank account, or a real human face all signal different audience clusters to the algorithm.

Agency owners selling to other agency owners need to show artifacts other agencies actually use. CRM screens, pipeline views, real Slack DMs from clients, real ads from past campaigns, never stock B-roll.

A creative that says “we help agencies book more calls” while showing generic stock footage of laptops gets routed to a generic audience. A creative that says the same thing while showing a real Calendly with 14 booked calls this week gets routed to people who recognize what 14 booked calls in a week looks like.

Visual specificity is the targeting cue. Stock B-roll is a $0 lookalike audience.

Lever 3: hook variety across the set

Within an ad set, Meta’s serving algorithm rewards creative diversity. If you upload three creatives that all open with “Are you tired of low ROAS?” you’ve given Meta one targeting signal repeated three times.

Upload three creatives that open with completely different hooks instead. A question, then a polarizing statement, then a number reveal. Now Meta has three different signals to test against three different audience pockets. The winner reveals which pocket is biggest.

The 30 hooks playbook we built for our clients tests across 10 different hook formats per creative cycle. We rotate them as the testing layer of the funnel, not as the brand layer.

If you want the full breakdown of which hooks are converting in 2026, I put together the 30 scroll-stopping hooks doc here.

How to test for Andromeda-fit creative

You can’t tell which creatives are Andromeda-fit by looking at them. You have to test in the wild.

Here’s the testing structure I run for our agency clients.

Phase 1: hook test (Week 1). Run 5-7 creatives with 5-7 different hooks against the same body content, the same offer, and the same CTA. The only thing changing is the first 3 seconds. ABO at $30/day per creative for 5 days. Kill anything below a 25% hook rate (people watching past 3 seconds). You’re not optimizing for CPA in this phase. You’re identifying which hooks Meta’s brain is willing to feed.

Phase 2: body test (Week 2). Take the 2-3 winning hooks from Phase 1 and test 3 different body structures behind each. Testimonial flow, problem-agitation flow, numbers-driven flow. CBO at $100/day for 7 days. Now you’re optimizing for cost per landing page view and cost per booked call.

Phase 3: scale (Week 3+). Move winners into a dedicated scaling campaign with broader audience and 2-3x daily budget. Watch CPA stability for 14 days before pushing harder.

Most agencies skip Phase 1 and go straight to scaling whatever they made first. That’s why their accounts plateau. Andromeda needs the hook layer dialed before it can scale anything.

What this means for budget allocation

The most expensive mistake post-Andromeda is under-investing in creative production.

Pre-2024, agencies spent maybe 10% of their ad budget on creative production. The other 90% went to media spend. That math worked when audience targeting did the heavy lifting.

Post-Andromeda the math is different. If creative is doing 60-70% of the targeting work, your creative production budget needs to scale with your media spend.

Our internal benchmark for agency clients running $5K-15K/mo on Meta is 25-30% of total budget on creative. That’s $1,250 to $4,500 monthly producing fresh creative iterations.

Underspend on creative and your $5K/mo media budget is converting at one-third efficiency. Overspend on creative and you’ve got assets sitting on the shelf nobody’s deploying.

The right ratio depends on your scale. Below $2K/mo on Meta the algorithm doesn’t have enough volume to learn from your creative variety, so the creative budget gets wasted. Above $2K/mo the math starts paying off. Above $10K/mo creative production becomes the lever that decides whether you 2x or plateau.

This is also why I tell agency owners not to start Meta ads until they can commit at least $2K/mo to media spend. Below that threshold Andromeda can’t work for you because there’s not enough data flowing through the pixel for the algorithm to find your audience via your creative signal.

The agencies winning right now are the ones who stopped treating creative as a brand asset and started treating it as a testing layer. They’re producing 8-12 creatives a month, killing the bottom 60%, scaling the top 20%, and iterating on the middle 20%.

That’s what Andromeda actually rewards.

The bottom line

Meta didn’t break between 2024 and 2026. It got smarter. The agencies blaming the algorithm are usually the ones still optimizing the audience side of the panel while ignoring the creative side.

Your creative is your targeting now. The first 3 seconds is your audience-selection moment, the visual specificity routes your delivery, and the hook variety inside an ad set decides how much Meta has to learn from.

Fix those three things before you touch another lookalike.

If you want a deeper look at the funnel architecture we run for agency clients booking 10-20+ qualified calls a month off Meta, read our 9-stage booking funnel breakdown, or grab the long-form self-optimizing funnel doc.

Want me to audit your setup?

If your CPA is creeping past $150 per booked call and you’re already on $5K+/mo Meta spend, the leak is almost always at the creative-targeting layer post-Andromeda.

I’m taking on 5 agency clients at a time. The last one we worked with 10x’d their sales to $300K+ in 90 days off the self-optimizing funnel. Brandon, mentioned earlier, dropped his CPA 86% in 2 weeks and generated $50K for his client in 10 days.

If you want me to look at your account directly and tell you exactly what’s leaking, book a 40-minute strategy call here. The audit is free. If we’re a fit we’ll talk about working together. If we’re not, you’ll walk out with the leak diagnosed.

Frequently asked questions