How do lookalike audiences work?
The process has three parts:- You provide a source audience. This is a Custom Audience built from your customer list, website visitors, or purchasers. Meta needs at least 100 people in the source, but 1,000+ works much better.
- Meta analyzes the source. The algorithm looks at hundreds of signals: demographics, interests, online behavior, purchase patterns, device usage, and more. It builds a profile of what your source audience “looks like.”
- Meta finds new people who match. It searches its 3+ billion users for people who resemble your source audience but aren’t already in it. You choose how closely they need to match by selecting a percentage.
What percentage should you use?
The percentage controls how closely the lookalike audience matches your source. Lower percentages are more similar but smaller. Higher percentages are broader but less precise.| Percentage | Audience Size (US) | Similarity | Best For |
|---|---|---|---|
| 1% | ~2.4 million | Highest match | Initial testing, small budgets ($20-50/day) |
| 2% | ~4.8 million | Very similar | Scaling after 1% works |
| 3% | ~7.2 million | Similar | Moderate budgets ($50-100/day) |
| 5% | ~12 million | Moderate match | Larger budgets, broader reach |
| 10% | ~24 million | Loosest match | High-spend accounts ($500+/day), awareness campaigns |
Start with 1%. If your CPA is good but you can’t spend your full budget (delivery is limited), move to 2-3%. Only go to 5-10% if you’re spending $200+/day and need more scale.
Lookalike audiences in plain English
Think of it like word-of-mouth referrals, but automated. Your best customers have friends, coworkers, and neighbors who share similar tastes, income levels, and shopping habits. If your best customer is a 32-year-old who shops online for premium skincare, Meta will find other people in that profile. You’re not picking interests from a dropdown and hoping. You’re telling Meta “find more people like these” and letting the algorithm do the pattern matching across billions of data points.Common lookalike audience mistakes
Source audience is too small
Source audience is too small
Meta requires at least 100 people in your source audience, but that bare minimum gives the algorithm very little to work with. With 100 people, Meta is building a profile from a tiny sample. Aim for 1,000-5,000 people in your source. If you don’t have enough purchasers yet, use add-to-cart events or email subscribers as your source instead.
Using the wrong source audience
Using the wrong source audience
A lookalike based on “all website visitors” includes tire-kickers, accidental clicks, and bots. A lookalike based on “customers who purchased 2+ times” gives Meta a clear signal of your ideal buyer. The quality of your source determines the quality of your lookalike. Use your best customers, not all customers. If you have purchase data, sort by LTV and use the top 25%.
Ignoring data freshness
Ignoring data freshness
A customer list from 2 years ago reflects who your customers were, not who they are now. People’s behavior changes. Meta’s user base changes. Refresh your source audiences every 30-60 days. If you’re using a website Custom Audience, set the lookback window to 30-90 days, not 180.
Not testing different percentages
Not testing different percentages
Many advertisers pick 1% and never test anything else. Run a simple split test: 1% vs 3% vs 5% with the same ad creative and budget. You might find that 3% delivers a lower CPA than 1% because the larger pool gives Meta more room to optimize delivery.
Stacking too many lookalikes in one campaign
Stacking too many lookalikes in one campaign
Running five lookalike ad sets at different percentages in the same campaign creates audience overlap. Meta ends up bidding against itself for the same people. Either use exclusions (exclude 1% from 2-3%, exclude 1-3% from 4-5%) or consolidate into a single broader lookalike and let CBO allocate budget.
How to create effective lookalike audiences
Build a high-quality source audience
Go to Meta Ads Manager, then Audiences, then Custom Audiences. Upload your customer list (email + phone for best match rates) or create an audience from your Meta Pixel events. Use purchasers or high-value customers, not all website visitors. A 1,000-person list of repeat buyers beats a 50,000-person list of page viewers.
Wait for the source to populate
After creating a Custom Audience, Meta needs time to match your data against its users. Customer list uploads typically match 50-70% of records. Wait until the audience status shows “Ready” before building a lookalike from it.
Create the lookalike audience
In Audiences, click “Create Audience” then “Lookalike Audience.” Select your source audience, pick your target country, and choose 1% to start. You can create multiple lookalikes at different percentages from the same source.
Test against other targeting methods
Run your lookalike audience alongside an interest-based ad set and a broad targeting ad set with the same creative and budget. Compare CPA and ROAS after 50+ conversions per ad set. In many accounts spending $100+/day, broad targeting now outperforms lookalikes because Meta’s algorithm has enough conversion data to find buyers on its own.
Refresh and iterate
Update your source audience monthly. As you get more customers, your lookalike quality improves. Test value-based lookalikes (weighted by purchase amount) versus standard lookalikes. And if you’re scaling spend, gradually increase the percentage rather than jumping from 1% to 10%.
Lookalike audiences vs other targeting types
| Targeting Type | How It Works | When It Wins |
|---|---|---|
| Lookalike (1-3%) | Meta finds people similar to your customers | Small-to-medium budgets, strong source data |
| Interest targeting | You pick interests/behaviors manually | New accounts with no pixel data, niche products |
| Broad targeting | No targeting restrictions, algorithm decides | High-spend accounts (50+ conversions/week), mature pixels |
| Retargeting | People who already visited/engaged | Bottom-funnel, re-engaging warm audiences |
See which audiences actually drive profit
AdAdvisor breaks down your ROAS and CPA by audience type so you can see whether your lookalikes are outperforming broad targeting or burning budget on low-quality traffic. Stop guessing which percentage works. See the numbers.Try AdAdvisor Free
Compare ROAS across lookalike, interest, and broad audiences in one dashboard.
ROAS Calculator
Calculate your break-even ROAS to know what “profitable” actually means for your business.