Strategy · 14 min read
Minting Marketing Team · Minting Marketing
Get my free auditLookalike audiences are Meta's most powerful targeting feature, but most advertisers use them wrong. They create one LAL from their website visitors and call it a day. Here's a deep dive into LAL strategy — from source audience selection to percentage sizing to layered testing — so you can unlock the full potential of this targeting tool.
A Lookalike Audience (LAL) is a targeting option that lets Meta find new people who share characteristics with an existing group of people you define (your "source audience"). Meta analyzes hundreds of data points — demographics, interests, behaviors, purchase patterns, device usage, content engagement — to find people who "look like" your best customers.
The key insight most advertisers miss: the quality of your Lookalike is entirely determined by the quality of your source audience. Garbage in, garbage out. A 1% LAL based on a poor source audience will underperform a 5% LAL based on an excellent source audience every single time.
This is where 90% of advertisers go wrong. They use their largest available audience as the source, assuming bigger is better. In reality, you want your source audience to represent your best customers, not all customers.
"We had been running a 1% LAL based on all website visitors for months with mediocre results. When we switched the source to our top 25% of customers by lifetime value, our cost per acquisition dropped 41% in the first two weeks. Same percentage, same budget — just a dramatically better source audience." — SaaS Company Marketing Lead
Meta recommends a minimum of 100 people in your source audience, but the sweet spot is 1,000-50,000. Below 1,000, Meta doesn't have enough data points to find meaningful patterns. Above 50,000, the audience becomes too diluted — it includes too many different "types" of people for Meta to identify what makes them similar.
When you create a Lookalike, you choose a percentage (1-10%). This determines how closely the new audience matches your source:
The answer depends on your budget and where you are in scaling:
Advanced advertisers don't just create one Lookalike — they create multiple LALs from different source audiences and test them against each other. This is "LAL stacking," and it's one of the most effective ways to find untapped audiences.
You'll almost always find that one LAL source significantly outperforms the others. That becomes your primary scaling audience while you continue testing new sources.
Most advertisers don't know this feature exists. When you upload a customer list as your source audience, you can include a "value" column (lifetime spend, AOV, etc.). Meta will then create a Lookalike that not only matches the characteristics of your customers but specifically prioritizes finding people similar to your highest-value customers.
We've seen value-based LALs outperform standard LALs by 25-40% on ROAS. Meta's algorithm focuses on finding people who look like your big spenders, not just your average customer.
Lookalike audiences aren't "set it and forget it." The source audience changes as you acquire new customers, and the broader population changes as people's behaviors evolve. Stale LALs lose their edge.
LALs can be combined with other targeting parameters. Whether you should layer additional targeting depends on your budget and audience size.
The general rule: at 1-2% LALs, let them run without interest layers. At 5%+, consider adding one broad interest layer to narrow the field.
There's a growing debate in the advertising world: are LALs still worth it, or has Meta's broad targeting algorithm gotten so good that you should just go broad? The honest answer: it depends.
Run both. Allocate 60% of prospecting budget to your best LALs and 40% to broad targeting. Compare performance monthly. Shift budget toward whichever performs better, but always keep both running — they often complement each other.
Create a LAL from your worst customers (high refund rate, low LTV, frequent complainers) and exclude it from your targeting. This tells Meta "don't find me people like these." Surprisingly effective for improving lead and customer quality.
Instead of basing LALs on customer lists or website visitors, use specific events: people who watched 75% of your best video ad, people who clicked "Get Directions" on a previous ad, people who saved your Instagram post. These event-specific behaviors signal strong intent.
If you're expanding to a new country, create a LAL using your home country customers as the source, targeted to the new country. Meta will find people in the new market who resemble your existing customer base. This is the fastest way to enter a new market with paid ads.
Lookalike audiences remain one of the most powerful tools in the Meta Ads arsenal — but only if you use them correctly. The difference between a mediocre LAL strategy and a great one isn't complexity. It's starting with the right source audience and systematically testing from there. Get the source right, and Meta's algorithm does the heavy lifting.
Lowest Cost gets you the most conversions for your budget. Cost Cap protects your margins but can choke delivery. Choosing wrong costs you money in both directions — here's the framework for deciding and the signals that tell you when to switch.
The old playbook was to stack interests, behaviors, and lookalikes to find the perfect audience. In 2026, going broad — age, gender, country only — often outperforms every detailed targeting option. Here's why it works and how to do it correctly.