Lookalike Audiences: Finding Your Ideal Customers on Facebook

Lookalike audiences let Meta take a list of people who already matter to your business and go find more people who resemble them across Facebook and Instagram. Done well, a Facebook lookalike audience can cut your cost per acquisition noticeably compared with broad interest targeting. Done lazily, it just burns budget on strangers who happen to share a few data points with your worst customers.

This post covers how lookalikes actually work, how to choose a seed list that is worth copying, and what has changed now that Meta pushes advertisers toward automated audience expansion.

What a Facebook Lookalike Audience Actually Is

A lookalike audience is a new group of users that Meta builds by analysing the shared characteristics of a source list you supply, then finding similar people in a country or region you pick. The source can be a customer file, website visitors captured by the Meta Pixel, app users, video viewers, page engagers or lead form submitters.

Meta does not tell you which signals it matched on, and you should not assume it is simply age and interests. The model weighs behavioural patterns, device use, purchase signals and hundreds of other variables that are never exposed in the interface.

The practical consequence matters: you cannot fix a bad lookalike by editing the audience. You fix it by changing the seed.

Lookalike Audience vs Custom Audience

People mix these up constantly, and the difference decides how you spend. A custom audience is made of people who already have a relationship with you, so it is retargeting territory. A lookalike audience is cold prospecting built on the shape of that relationship.

• Custom audience: your email list, past purchasers, visitors to a pricing page, people who watched 75% of a video.
• Lookalike audience: strangers who statistically resemble one of those custom audiences.
• Budget logic: custom audiences usually convert cheaper but cap out fast, while lookalikes scale but need patience.

A sensible account runs both, with exclusions in place so your prospecting campaigns are not paying to reach people who bought last week.

The Part Most Guides Skip: Seed Quality

Almost every tutorial tells you to click “create audience” and pick 1%. Very few talk about the thing that decides whether the campaign works, which is the quality and size of the list you feed in. Meta’s own guidance suggests a source of at least 1,000 people from a single country, and in practice 2,000 to 50,000 of the right people outperforms 200,000 mediocre ones.

Before you build anything, score your seed list against a few questions:

• Are these buyers or browsers? A list of purchasers beats a list of everyone who visited your homepage.
• Is it recent? Customers from the last 90 to 180 days reflect your current offer and pricing better than a four year old export.
• Is it clean? Match rates climb when you include email, phone, first name, last name, city and ZIP rather than email alone.
• Is it concentrated? One country per source audience, because mixing markets muddies the model.
• Does it exclude refunds and churn? Copying people who asked for their money back is an expensive way to find more of them.

If your best seed has fewer than 1,000 records, build a lookalike from a strong engagement signal instead, such as people who spent time on a product page or completed a lead form.

Value-Based Lookalikes Are the Upgrade Nobody Uses

If you upload a customer file with a value column (lifetime value, total spend or margin), Meta weights the model toward your highest value customers rather than treating every row equally. This is the single biggest improvement most accounts can make in an afternoon.

Pull the last 12 to 24 months of transactions, sum revenue per customer, strip outliers such as one wholesale order that dwarfs everything, and upload that. Advertisers often see meaningfully better return on ad spend from a value-based seed than a flat list of the same size, because the top decile of customers rarely looks like the average one.

For a closer look at this topic, see our guide: Using Looker Studio (formerly Data Studio) for Custom PPC Reporting.

Related reading: Evergreen Content vs. Trending Content: Finding the Sweet Spot.

Related reading: Best Practices for Facebook Ad Creative in 2026.

Related reading: Retargeting on Facebook: Strategies for Insane ROI.

How to Create a Lookalike Audience in Facebook Ads

The mechanics take about five minutes once your source exists. In Meta Ads Manager:

1. Open Audiences from the main menu, then choose Create Audience and select Lookalike Audience.
2. Pick your source: an existing custom audience, a customer file with values, your Pixel, or a Page.
3. Choose the location, which should be the country or countries you actually sell and ship to.
4. Set the audience size using the percentage slider, from 1% up to 10% of that country’s population.
5. Create several sizes at once (1%, 3%, 5%) so you can test them in separate ad sets later.
6. Wait 6 to 24 hours for the audience to populate before launching, and check refresh settings if the source updates automatically.

Meta’s documentation in the Business Help Center is worth a read if you need the current field requirements for customer file uploads, since the hashing and formatting rules change periodically.

Which Percentage Should You Choose?

The slider is not a quality dial so much as a precision-versus-reach trade. A 1% lookalike in the United States is roughly 2.5 million people, which sounds huge but is tight by Meta standards.

• 1%: closest match, best for small budgets, high ticket offers or narrow niches.
• 2% to 5%: the usual sweet spot once a campaign is spending consistently and the 1% starts to fatigue.
• 6% to 10%: reach plays for broad consumer products with cheap price points and lots of creative.

Start narrow, prove the economics, then widen. If your 1% cannot make the numbers work, a 10% version will not save it.

Are Facebook Lookalike Audiences Going Away?

This question comes up a lot, and the short answer is that they still exist but they are no longer the default. Meta now steers advertisers toward Advantage+ audience, where your lookalike or custom audience becomes an “audience suggestion” that the system can spend outside of when it finds better results elsewhere.

In practice that means your carefully built lookalike acts as a starting hint rather than a hard boundary. Some accounts see better cost per result from that freedom, others lose control of who they reach, so the only honest approach is to test both setups with the same creative and budget.

Keep building lookalikes regardless. Even when the algorithm expands past them, a strong seed teaches the system faster than launching with no signal at all. If you want the wider picture on structuring targeting across campaign types, our guide on mastering Facebook Ads targeting strategies covers how these pieces fit together.

Privacy Changes and Why Your Match Rates Dropped

Since Apple’s App Tracking Transparency rollout, a large share of iOS users decline app tracking, which shrinks the signal flowing back to Meta. Facebook custom audiences built purely from Pixel events are smaller and slower to refresh than they were in 2020, and that weakens the lookalikes built on top of them.

Two fixes matter more than the rest. Install the Conversions API so server-side events supplement browser data, and lean harder on first party customer files that do not depend on browser tracking at all. Meta’s own reporting is clear that hashed customer lists remain one of the more durable signals available.

Handle that data properly. Upload only information you collected with permission, honour opt-outs, and keep your privacy policy aligned with the FTC’s privacy and security guidance for marketing data.

Testing Structure That Does Not Waste Budget

Lookalikes fail more often from bad testing than bad audiences. Give each one a fair shot with enough budget to exit the learning phase, which generally means around 50 optimisation events per ad set per week.

• Run one variable at a time: same creative across three lookalike sizes, or same size across three seeds.
• Exclude existing customers and recent purchasers from every prospecting ad set.
• Avoid stacking overlapping lookalikes in one campaign, since they compete in the same auction.
• Refresh seed lists every 30 to 90 days so the model tracks your current buyer, not last year’s.
• Judge on cost per purchase or qualified lead, never on click-through rate alone.

The same discipline applies on the search side. If you run both channels, the logic behind custom intent audiences in Google Ads mirrors this approach closely, and setting up your first Facebook campaign is worth revisiting if your account structure is still messy.

Frequently Asked Questions

What are lookalike audiences on Facebook?

A Facebook lookalike audience is a targeting group Meta generates by finding users who share characteristics with a source list of at least 1,000 people you provide. The source can be customers, website visitors, app users or people who engaged with your Page, and Meta then finds similar users within a country you select.

Can you give me an example of a lookalike audience?

A common example is a 1% lookalike built from the last 5,000 purchasers of an online store, which in the United States reaches roughly 2.5 million people. A stronger version uploads the same customers with their lifetime value attached, so Meta prioritises finding people who resemble the top spenders rather than the average buyer.

What are the four types of target audiences?

Most advertisers work with four groups: core or interest-based audiences, custom audiences of existing contacts, lookalike audiences built from those customers, and broad or Advantage+ audiences where the algorithm chooses. Each serves a different stage, with custom audiences handling retargeting and lookalikes handling cold prospecting at scale.

How do you set up a lookalike audience on Facebook?

Go to Audiences in Ads Manager, click Create Audience, choose Lookalike Audience, select your source, pick a country and set the size between 1% and 10%. The audience typically takes 6 to 24 hours to populate before it can be used in an ad set, so build it a day before launch.

Want a Second Opinion on Your Facebook Targeting?

If your lookalikes are spending without converting, the problem is usually the seed list, the exclusions or the offer rather than the slider. Talk to SEO Quirk about a review of your account, or read why working with a local agency tends to shorten the feedback loop on campaigns like these.