How to build a lookalike audience strategy that still works

The old lookalike playbook is broken

You used to be able to upload a list of your best customers, create a 1% lookalike, and watch Meta print money. That stopped working reliably around 2022. iOS 14.5 degraded match rates. Audience sizes shrank. Many brands responded by going broad (Advantage+ with no targeting), and some of them did fine. But plenty of others saw CPAs spike and wondered what happened.

Lookalikes aren't dead. They just need a different approach. Here's what actually works now.

Why lookalikes degraded and why they still matter

Two things changed at once. First, Apple's ATT prompt reduced the signal Meta gets from your pixel. Fewer events matched back to users means your seed audiences got noisier. Second, Meta's algorithm got better at broad targeting, so the gap between a good lookalike and broad narrowed.

But lookalikes still have a role. They give Meta's delivery system a starting point. When you're spending $200 to $500 a day (not $5,000), that starting point matters. Broad works best when you have massive creative volume and big budgets. For most D2C brands in the $50K to $300K/month spend range, a smart lookalike strategy still outperforms pure broad on a per-ad-set basis.

Start with better seed audiences

The seed audience is everything. Most brands default to "all purchasers" or "website visitors." These are fine but not great. Here's what to use instead:

• High-LTV customers. Export your top 25% of customers by lifetime spend from Shopify or your CDP. This is a fundamentally different signal than "everyone who bought once."
• Repeat purchasers only. If you sell consumables or have a subscription, a list of people who bought 2+ times tells Meta to find people who stick around, not just impulse buyers.
• High-AOV purchasers. If your average order is $65 but some customers spend $150+, isolate them. You want Meta to find more of those people.
• Email engagers who purchased. Pull contacts from Klaviyo who opened 3+ emails AND purchased. This filters out discount hunters and passive subscribers.
• Video viewers (75%+ completion). This is a platform-native signal, so it doesn't rely on pixel matching. If you have videos that explain your product well, 75% viewers make a strong seed.

Keep seed lists between 1,000 and 10,000 people. Smaller than 1,000 and Meta doesn't have enough data. Larger than 10,000 and you're diluting the signal.

How to structure the ad sets

Don't just create one lookalike and call it done. Test these variables:

• Percentage ranges. 1% is the tightest. 1 to 3% is usually the sweet spot for most D2C brands. 5% and above starts to look like broad. Test 1%, 1 to 3%, and 3 to 5% separately.
• Stacking vs. isolating. You can combine multiple seed audiences into one lookalike (stacking) or run them separately. Stacking gives Meta more flexibility. Isolating lets you see which seed performs best. Start isolated, then stack your winners.
• Country targeting. If you sell in the US only, your lookalike is US-only. Obvious, but people forget to check this when duplicating campaigns.

Run each lookalike ad set with the same 3 to 4 creatives so you're comparing audiences, not ads.

Refresh your seeds every 60 to 90 days

This is the step most brands skip. Your customer base changes. Seasonality shifts who buys. A seed list from Q4 is full of gift buyers who may not represent your core customer. Export fresh lists quarterly at minimum. Monthly is better if you have the volume.

Set a calendar reminder. It takes 15 minutes and can meaningfully change performance.

When to use lookalikes vs. broad vs. Advantage+

Simple framework:

• Lookalikes work best for prospecting when your budget is moderate and your creative volume is limited (under 10 new ads per week).
• Broad targeting works when you have high creative volume and strong pixel data from consistent spend.
• Advantage+ Shopping Campaigns (ASC) blend both. They work well but give you less visibility into what's driving results. Use them alongside manual lookalike campaigns, not as a replacement.

The brands we see winning run all three and compare. There's no universal answer.

Keep testing the inputs

The real advantage isn't in finding one perfect lookalike. It's in systematically testing seed audiences, refreshing them, and pairing them with the right creative. That combination is what separates accounts that scale from accounts that stall.

At Argus, we build and maintain lookalike strategies as part of our paid media management. We pull the seed lists, test the structures, and refresh everything on a schedule so performance doesn't quietly decay. If your Meta account feels stuck on the targeting side, ask for a free growth plan and we'll show you what we'd change.