How to Build Meta Lookalike Audiences That Convert

What a Lookalike Audience Actually Is

Meta takes a list of your existing customers, finds the shared traits in that list (not just demographics, but behavior patterns Meta can see across its platform), and then builds a new audience of people who share those traits. That new group is your lookalike audience. The core promise is simple: you stop guessing who might buy from you and start targeting people who already resemble the people who did.

The catch is that most advertisers set these up once, let them run, and never touch them again. A lookalike audience is not a "set it and forget it" targeting option. It's only as good as the source data you feed it, and it needs regular maintenance to stay useful as your customer base changes.

In our experience, the difference between a lookalike that converts and one that just burns budget almost always comes down to two things: what source audience you used to build it, and how tightly you defined the percentage match. We'll walk through both.

[Ad space]

The Three Inputs That Make or Break Your Lookalike

Before you open Ads Manager, you need to think through three decisions. Get these wrong and no amount of ad creative will save the campaign.

1. Source audience quality. Meta needs a source list to model from. Your options typically include:

  • Customer list (uploaded via email or phone number)
  • Website visitors (via the Meta Pixel)
  • People who engaged with your Facebook or Instagram page
  • App users
  • Purchasers specifically, pulled from pixel events

A list of actual purchasers is almost always stronger than a list of "everyone who visited the site." Visitors include window shoppers, people who bounced in two seconds, and bots. Purchasers are people who already pulled out a credit card for you. If you only have one good source, make it your buyer list.

2. List size. Meta recommends a source list of at least a few hundred people, but in our experience the quality jumps noticeably once you're feeding it 1,000 or more real customers. Too small, and Meta doesn't have enough pattern to work with. Too broad (like your entire email list going back five years, including people who never bought anything), and you dilute the signal.

3. Match percentage. When you build the lookalike, Meta asks you to choose how closely the new audience should match your source, usually on a 1 to 10 scale where 1% is the tightest match and 10% is the broadest. A 1% lookalike in the US is typically around two million people who most closely resemble your source. A 10% lookalike might be closer to twenty million people who still resemble your source, just less precisely.

Step-by-Step: Building Your First Lookalike Audience

Here's the process in order, written so you can follow it inside Ads Manager without guessing.

  1. Go to Audiences in Meta Ads Manager and click Create Audience, then Lookalike Audience.
  2. Choose your source. If you have a customer list with at least a few hundred purchasers, upload that. If not, use pixel purchase events or, as a fallback, add-to-cart events.
  3. Pick the location you want the lookalike built from. If you sell only in Canada, don't build a US-based lookalike.
  4. Set the audience size. Start at 1% for your first test. This gives you the tightest, highest-quality match before you start broadening.
  5. Let it populate. Meta usually takes anywhere from a few hours to a day to build the audience, depending on source size.
  6. Exclude your existing customers from the new campaign targeting this lookalike. Otherwise you're paying to "acquire" people who already bought from you.
  7. Launch with a modest budget and let it run long enough to exit the learning phase before judging results. Changing the audience or budget too early resets this process.

A Simple Naming Template

Keep your audiences organized from day one. A naming convention like this saves hours later:

LAL 1% - Purchasers 90d - US - [Date Created]

This tells you at a glance what percentage match it is, what source it came from, the time window, the country, and when you built it. When you have ten lookalikes running, this detail matters.

Common Mistakes That Kill Lookalike Performance

Most underperforming lookalike campaigns trace back to one of these issues:

  • Using a tiny or stale source list. A customer list from three years ago may not reflect who buys from you today, especially if your product or pricing has changed.
  • Building a lookalike from all website visitors instead of buyers. This usually produces a broader but weaker match.
  • Going straight to 10% broad match. Broader audiences are cheaper to reach but typically convert at a lower rate. Start tight, then expand only if you need more volume.
  • Never excluding existing customers. This wastes spend on people who don't need to be "acquired" again.
  • Combining too many lookalikes with overlapping sources in the same ad set. This can cause Meta's delivery system to compete against itself, raising costs for no real targeting benefit.
  • Judging results too fast. A campaign needs enough spend and time to exit the learning phase. Cutting it after two days tells you almost nothing.

How to Refine a Lookalike Audience Over Time

Once your first lookalike is running, treat it like a living asset, not a one-time setup task.

Refresh the source list regularly. If you're using a customer list, re-upload it every month or quarter with your latest buyers. If you're using pixel-based purchase events, Meta updates this automatically, but it's still worth checking that your pixel is firing correctly and that purchase volume is steady.

Test match percentages side by side. Run a 1% and a 3% version of the same source audience in separate ad sets with the same budget. Compare cost per purchase after each has had a fair run. This tells you whether tightening or loosening the match helps your specific offer.

Segment your source audience by value, not just by purchase. A lookalike built from your highest-spending customers often performs differently than one built from everyone who ever bought anything, including a single clearance item. If you have the data, try building a lookalike from customers who placed a second or third order. Repeat buyers tend to model better than one-time buyers.

Watch frequency and overlap. If the same lookalike audience has been running for months, check your frequency metric. Rising frequency with falling conversion rate is usually a sign the audience is fatigued and needs a refresh or a pause.

Don't abandon a lookalike after one bad week. Short-term noise happens. Look at trends over two to four weeks before deciding an audience isn't working.

Quick Troubleshooting Guide

If your lookalike campaign isn't converting, check these in order:

  1. Is the pixel still firing correctly on your site?
  2. Is your source list recent and made of actual buyers, not just visitors?
  3. Did you exclude existing customers from this campaign?
  4. Has the campaign had enough budget and time to exit the learning phase?
  5. Is the match percentage too broad for your budget size?

Most performance problems trace back to one of these five questions.

Actionable Summary

  • Build your first lookalike from actual purchasers, not general website visitors, whenever you have enough data to do so.
  • Start tight at a 1% match before testing broader percentages.
  • Exclude existing customers from every lookalike campaign targeting new buyers.
  • Refresh your source list monthly or quarterly so the audience reflects who buys from you now.
  • Give every test at least two to four weeks of real data before judging it.
  • Keep a simple naming convention so you can track which audience, percentage, and date each campaign is pulling from.

A lookalike audience is a tool, not a strategy on its own. Feed it good data, check it regularly, and give it time to work before you pull the plug.