Lookalike Audience

A targeting method where ad platforms find new people who resemble your best existing customers. You provide a "seed" audience (your buyers, your email list, your highest-value users), and the platform's algorithm scans its user base to find people with similar demographics, interests, and online behavior.

It's the closest thing to a shortcut in paid acquisition. Instead of guessing which interests to target, you let the algorithm figure out what your customers have in common and find more people like them.

1. You provide a seed audience. This is usually a custom audience: your customer list, your website converters, your highest-LTV buyers. The better the seed, the better the lookalike.

2. The platform analyzes your seed. Meta looks at hundreds of data points: age, location, interests, purchase behavior, device usage, engagement patterns. It identifies the things your best customers share.

3. The algorithm finds matches. It scans the broader user base and builds a new audience of people who score high on similarity to your seed.

On Meta, you choose how closely the lookalike should match your seed:

• 1% — the top 1% most similar people in a given country. Smallest audience, highest quality. Best for conversion campaigns.
• 3% — a good middle ground when you need to scale beyond 1% but still want quality.
• 5–10% — largest reach, lowest similarity. Better for awareness campaigns or when your 1% audience is tapped out.

A 1% lookalike in the US is roughly 2.5 million people. That's usually more than enough to run campaigns for months.

The algorithm is only as good as the data you give it. Here's a rough quality ranking for seed audiences:

1. 1.Paying customers (especially high-LTV ones)
2. 2.Trial-to-paid converters (people who evaluated and decided to buy)
3. 3.Key activation event completers (people who took a meaningful product action)
4. 4.High-intent page visitors (pricing page, checkout page)
5. 5.Email lists (buyer segments preferred over full lists)

Using your "all website visitors" as a seed produces a mediocre lookalike because that audience includes everyone from serious buyers to people who bounced after two seconds.

You have a proven converting audience as your seed. If you don't yet know who your ideal customer is, lookalikes won't fix that. You need real conversion data first.

Your retargeting pools are running low. Retargeting audiences are limited in size. Once you've reached everyone who visited your site, you need cold traffic. Lookalikes bridge the gap between retargeting and broad, untargeted prospecting.

You want to scale without manual interest testing. Instead of trying every combination of interest targets, let the algorithm do the pattern matching. It's usually better at it than you are.

Meta recommends a minimum of 100 people in your seed audience, but 1,000–10,000 is where you'll see the best results. Ideally your pixel should be recording 50+ conversions per week for the algorithm to learn well.

iOS 14+ changed things. Privacy changes reduced the accuracy of pixel-based lookalikes. First-party data (customer email lists, CRM data) has become more important as a seed source because it's not affected by tracking restrictions.

Lookalikes refresh every 3–7 days. The audience composition updates regularly, so performance can shift over time.

Audience overlap is real. If you run multiple lookalike audiences at the same time, they may contain many of the same people. Use Meta's audience overlap tool to check.

They won't fix a broken funnel. Lookalikes find people similar to your customers. If your landing page doesn't convert or your offer is weak, sending more of the "right" people there won't help.