Facebook Lookalike Audiences: How to Build and Optimize Them
Facebook Lookalike Audiences: How to Build and Optimize Them
Lookalike audiences let you hand Meta a list of your best customers and say: "Find more people who look like these." When built correctly, lookalikes routinely outperform interest-based targeting, especially at scale. When built carelessly, they dilute into noise. The difference between a high-performing lookalike and a wasteful one almost always comes down to the seed audience: the source data you give Meta to model from. This guide covers everything from seed selection to percentage choices to the post-iOS reality of lookalike performance in 2025.
What Lookalike Audiences Are and How Meta Builds Them
A Lookalike Audience is a targeting option that finds new people on Facebook and Instagram whose characteristics resemble those of a source audience you define. Meta's algorithm analyzes hundreds of signals from your seed audience, including demographics, interests, behaviors, and on-platform engagement patterns, then finds users in your target geography who match those patterns.
The algorithm does not clone your customers. It finds probabilistic matches across Meta's population of approximately 3 billion monthly active users. The larger and higher-quality your seed audience, the more signals Meta has to model from, and the more precise the resulting lookalike.
Meta officially states that a seed audience of 1,000 to 50,000 people produces the best results. Below 100 people, the model becomes unreliable. Above 50,000, you are often diluting your signal with lower-quality entries.
The Seed Audience Matters Most
The quality of a lookalike audience is a direct function of the quality of the seed. This is the most important principle in lookalike targeting, and it is where most advertisers make their biggest mistakes.
Customer Purchase Lists (Best Seed)
Uploading a list of actual purchasers, especially high-LTV customers, gives Meta the strongest signal. These are people who already converted, which means Meta can model on the behaviors and characteristics of people who actually buy, not just people who clicked or visited.
For best results, upload your top purchasers rather than all customers. If you sell a product with strong repeat purchase rates, segment your upload to customers who have bought three or more times. This tells Meta to find buyers who are likely to become loyal customers, not just one-time purchasers.
Pixel Events (Strong Seed)
If you do not have a large customer list, pixel events are your next best option. Create a custom audience from your "Purchase" pixel event over the last 180 days. If purchase volume is low (under 200 events), step up the funnel to "InitiateCheckout" or "AddToCart" and use those as seed audiences.
The key: always use the most valuable pixel event that gives you at least 1,000 people in the source audience. A purchase-based lookalike from 300 people will typically outperform an add-to-cart lookalike from 5,000 people.
Video Viewers (Scalable Seed)
For brands with video content, video engagement audiences are a scalable seed option. Create a custom audience of people who watched 75% or more of a video. High video completion indicates strong intent and interest: Meta can build a reliable lookalike from people who demonstrated that level of engagement.
This seed is particularly useful for brands without large customer lists who are investing in upper-funnel video content.
What Not to Use as a Seed
Avoid seeding from all website visitors (too broad, includes accidental traffic), email newsletter lists without segmentation (low-intent subscribers dilute your seed), or engagement-based audiences built from post likes and reactions (engagement does not correlate strongly with purchase intent).
Lookalike Percentages Explained
When creating a lookalike, you choose a percentage from 1% to 10%. This percentage refers to what portion of the target country's population Meta will include in your lookalike.
1% Lookalike
The most similar 1% of the population to your seed. Smallest audience, highest relevance. For the United States, a 1% lookalike is approximately 2.1 million people. Use 1% when you want the tightest match: it typically delivers the best conversion rates but limits scale.
2% to 5% Lookalike
Progressively larger and less targeted. The 2-5% range is a common starting point for prospecting campaigns with moderate budgets. You are trading some precision for scale. For accounts spending $3,000 to $15,000 per month on prospecting, 2-3% lookalikes often hit the best CPA-to-volume tradeoff.
6% to 10% Lookalike
Approaching broad targeting in terms of accuracy. The 10% lookalike for the US includes approximately 21 million people and only loosely resembles your seed. These larger percentages can make sense for very broad-appeal products or when you need significant scale and broad targeting is already working in your account.
A Common Testing Approach
Many experienced Meta advertisers run 1%, 2-3%, and 5-7% lookalikes as separate ad sets to find which percentage delivers the best CPA. Start with 1% and 3%, let both run for two weeks with equal budget, then double down on the winner.
How Many Lookalikes to Run at Once
Running too many lookalikes creates audience overlap, budget fragmentation, and muddied data. A practical limit for most accounts is 3 to 5 lookalike ad sets simultaneously within a prospecting campaign.
If you are using Campaign Budget Optimization (CBO), Meta will allocate budget dynamically across your lookalike ad sets, which handles some of the fragmentation risk. With manual budget per ad set, be more conservative: 2 to 3 lookalikes is usually enough.
Use Meta's Audience Overlap tool (under Audiences in Ads Manager) to check that your lookalike ad sets are not heavily overlapping with each other or with your retargeting audiences.
Lookalike Audiences in 2025: Do They Still Work After iOS Changes?
The honest answer: lookalikes are less precise than they were in 2021, but they still work and often outperform interest-based alternatives.
iOS 14.5's App Tracking Transparency (ATT) prompt, released in April 2021, caused a significant portion of iOS users to opt out of tracking. This reduced the volume of purchase events flowing into Meta's pixel from iOS users, which in turn reduced the quality of pixel-based seed audiences for some advertisers.
The practical impact depends heavily on your customer base. If your customers are primarily iOS app users and you relied on pixel data for seeding, your lookalikes lost accuracy. If your customers complete purchases on the web (especially on Android devices or desktop), the impact is more limited.
According to Optily's analysis, the effect on lookalike audiences themselves is more moderate than the effect on tracking and attribution. Meta's modeling has adapted significantly since 2021, and the platform's Conversions API (CAPI) allows brands to send server-side purchase data directly to Meta, bypassing some of the browser-level signal loss.
For 2025, the recommended approach is to use both pixel events and customer lists as complementary seed sources, implement Conversions API alongside your pixel, and supplement lookalike targeting with broad targeting and Advantage+ Audience as a parallel test.
Combining Lookalikes with Interest Targeting
Stacking lookalikes on top of interest targeting was a common tactic to narrow audience size. In 2025, Meta's official guidance leans against this. Adding detailed interest targeting to a lookalike often restricts the audience too much and limits the algorithm's ability to optimize.
The exception: for very niche products or small markets where your 1% lookalike is too small (under 300,000 people), adding one broad interest layer can be a practical workaround. Keep the interest broad rather than specific.
How Adwise Monitors Lookalike Performance and Recommends Shifts
Lookalike performance does not stay static. Audiences saturate, pixel data quality shifts with platform changes, and new customer segments emerge over time. Monitoring lookalike health manually requires regular audience performance breakdowns, saturation checks, and cost-per-result trend analysis.
Adwise tracks your Meta Ads account daily, including lookalike audience performance trends, CPM inflation signals (a sign of audience saturation), and comparative performance across your audience types. When a lookalike begins to underperform, Adwise surfaces that insight the next day with a specific recommendation: whether to refresh the seed, expand the percentage, or shift budget to a different audience type.
That daily monitoring replaces what otherwise takes 2 to 3 hours of weekly manual analysis for most advertisers managing accounts with multiple active audience strategies.
Build Better Lookalikes Starting Today
Lookalike audiences are one of the highest-leverage targeting tools on Meta Ads, but only when built on strong seeds, managed with the right percentage strategy, and monitored consistently over time. The brands that get this right scale efficiently. The ones that set it and forget it waste budget on degraded audiences.
Adwise monitors your Meta Ads account daily and tells you exactly which audiences are underperforming and what to do about it. Connect in 60 seconds, no credit card required.