What Is a Lookalike Audience?
A targeting audience built from new users who behave like your existing high-value users.
Book a free callA lookalike audience is a targeting audience made up of new users who resemble a seed audience you give the platform — for example, your paying users — in behavior and interests. The algorithm learns the shared traits of the seed audience and finds people who fit the same profile. The quality of the seed audience directly determines the quality of the result.
How do you build one?
- Choose the seed audience. This is the most critical decision. Instead of "all users," pick a value-generating segment such as "purchasers in the last 90 days."
- Make sure it's big enough. Platforms usually require a seed of at least 1,000 people; an audience of 1,000-5,000 works well in most cases.
- Set the similarity percentage. 1% is the narrowest and most similar match; 10% is the broadest and loosest.
- Define the geography. Lookalike audiences are generated per country.
- Refresh the seed regularly. A stale seed audience doesn't reflect your current user profile.
How do you choose the similarity percentage?
| Percentage | Audience size | When to use it |
|---|---|---|
| %1 | Smallest | When you need high precision, during testing |
| %2 – %3 | Medium | Balance of quality and scale; the most common choice |
| %5 – %10 | Largest | During scaling, when you need volume |
A practical approach: start at 1% to validate performance, then widen the percentage gradually as you need more scale. As the percentage grows, audience quality usually drops but CPM may drop too; always test the net effect.
Why does seed audience selection matter so much?
The lookalike algorithm imitates the seed audience. If you feed it "everyone who installed the app," the algorithm also learns from people who installed and churned right away and brings you more of them.
- Best seeds: High spenders, repeat purchasers, long-term active subscribers.
- Middle ground: Users who signed up or completed a specific onboarding step.
- Weak seeds: All installs, all site visitors, your entire email list.
If a value-based lookalike option is available, use it: when you also send the spend value for each user in the seed audience, the algorithm prioritizes not just similar users but high-value lookalikes.
Common mistakes
- Using a seed that's too broad. The more generic the seed, the more ordinary the lookalike; eventually it's no different from broad targeting.
- Never updating the seed. A seed audience created months ago doesn't reflect your current product and user profile.
- Not using exclusions. Your existing users can end up in the lookalike audience; if you don't exclude them, you spend budget showing ads to users you already have.
- Building heavily overlapping audiences. Lookalike audiences that largely overlap bid against each other in the auction and drive up costs.
- Relying only on lookalikes. In modern campaign structures, broad targeting plus a strong conversion signal often outperforms narrow lookalike audiences. Test both.
Frequently asked questions
How big should the seed audience for a lookalike be?
Platforms' minimum is usually 1,000 people, but what determines quality isn't size, it's homogeneity. A well-defined "high spenders" segment of 2,000 people delivers far better results than an "all installs" list of 50,000.
Are lookalikes still effective?
Less so than they used to be. Privacy restrictions have reduced signal quality, and platform algorithms have become markedly better at broad targeting. Today, in most accounts, broad targeting combined with the right conversion event works at least as well as lookalikes. Still, they're a tool worth testing.
What difference does a value-based lookalike make?
A standard lookalike finds people who resemble the seed audience; a value-based one gives more weight to the high spenders within the seed. As a result, the users you acquire have a higher average LTV. If you can send purchase value data, it should almost always be your choice.
Let's update your audience strategy
We steer your campaigns toward high-LTV users with value-based audiences and the right conversion signals.
Book a free call