Facebook Lookalike Audiences: Find More Buyers Like Your Best Customers

What Are Lookalike Audiences?

Lookalike audiences are one of Meta's most powerful targeting features. You give Meta a list of your existing customers or high-value website visitors, and the algorithm finds millions of similar people who share the same characteristics, behaviors, and interests.

Think of it as cloning your best customers. Instead of guessing who might buy your product through interest targeting, you let Meta's machine learning find people who statistically resemble the people who already bought from you.

When done right, lookalike audiences consistently outperform interest-based targeting by 30-60% on cost per acquisition.

How Lookalike Audiences Work

Meta analyzes your source audience across hundreds of data points including demographics, interests, online behavior, purchase patterns, device usage, and engagement history. Then it scans its entire user base to find people who match that profile.

The result is a new audience of people who have never interacted with your brand but share key traits with people who have. These are statistically the most likely people to become your next customers.

Building Your Source Audience

The quality of your lookalike depends entirely on the quality of your source audience. Better source data produces better lookalikes.

Best Source Audiences (Ranked)

1. Purchase event data (best): People who actually bought from your store. This is the gold standard because it tells Meta exactly what a buyer looks like.
2. Add-to-cart event data: People who added products to their cart. Not as strong as purchases but still shows high intent.
3. Customer email list: Upload your customer emails and Meta matches them to profiles. Effective when you have 500+ emails.
4. Website visitors (all): Everyone who visited your store. Broad but still valuable as a starting point.
5. Video viewers (75%+): People who watched most of your video ad. Shows strong interest in your product category.
6. Page/profile engagers: People who interacted with your Facebook or Instagram content. Weakest signal but useful when other data is limited.

Minimum Source Size

Meta requires at least 100 people in your source audience, but aim for 500-1,000+ for best results. The more data points Meta has, the better it can identify patterns.

The critical threshold: You need approximately 50 purchase events before a purchase-based lookalike becomes truly effective. Below that, the sample size is too small for Meta's algorithm to find reliable patterns.

Choosing Your Lookalike Percentage

When creating a lookalike, you choose a percentage that determines how closely the new audience matches your source:

• 1% lookalike: The top 1% of the population most similar to your source. Smallest audience, highest match quality. In the US, this is roughly 2.1 million people.
• 2-3% lookalike: Slightly broader. Still strong match quality with a larger reach. Good balance of precision and scale.
• 5% lookalike: Much broader. Lower match quality but significantly more reach. About 10.5 million people in the US.
• 10% lookalike: Very broad. Approaching broad targeting territory. Only useful for scaling with large budgets.

Which Percentage to Start With

Start with 1%. It almost always outperforms broader percentages for new campaigns. Once you have exhausted the 1% audience (frequency creeping above 2.0), expand to 2-3%.

Testing Multiple Percentages

Create separate ad sets for 1%, 3%, and 5% lookalikes and let them compete. You will typically see the 1% win on CPA, the 3% deliver decent CPA with more volume, and the 5% sacrifice efficiency for scale.

Advanced Lookalike Strategies

Stacking Lookalikes

Combine multiple source audiences into one lookalike for broader but still targeted reach:

• Create a 1% lookalike from purchasers
• Create a 1% lookalike from add-to-carters
• Combine them in one ad set using OR logic (target people in either audience)

This gives the algorithm more room to optimize while keeping quality high.

Value-Based Lookalikes

If you pass purchase values to your pixel, you can create value-based lookalikes. Meta prioritizes finding people who resemble your highest-spending customers, not just any customer.

This is particularly powerful if you sell products at different price points. Meta will optimize for finding big spenders rather than bargain hunters.

Excluding Existing Audiences

Always exclude your existing customers and recent website visitors from lookalike campaigns. You do not want to pay to reach people who already know about you. Use custom audience exclusions at the ad set level.

Refreshing Source Audiences

Your customer base evolves over time. A lookalike built from customers six months ago may not reflect your current buyer profile. Use dynamic source audiences that automatically update:

• Website custom audiences with time windows (last 30, 60, or 90 days)
• Pixel events that continuously add new data points
• Regularly re-upload customer lists if using email-based sources

Lookalike Audiences in Your Campaign Structure

Testing Phase (Pre-Lookalike)

When you first launch a store, you do not have enough data for lookalikes. Run interest-based campaigns to generate initial sales and pixel data.

Goal: Accumulate 50+ purchase events as quickly as possible. This is your ticket to creating effective lookalikes.

Transition Phase

Once you hit 50 purchases, create your first 1% purchase-based lookalike. Run it alongside your best-performing interest-based ad sets and compare performance over 7 days.

In most cases, the lookalike will match or beat interest targeting on CPA within the first week.

Scaling Phase

As your pixel accumulates more data (100, 500, 1,000+ purchases), your lookalikes become increasingly powerful. At this stage:

• Lookalikes should be your primary prospecting audience
• Interest targeting becomes secondary for testing new angles
• Layer broader lookalikes (3-5%) as you increase budget
• Continue feeding the pixel with new conversion data

Common Mistakes to Avoid

Using Too Small a Source Audience

A source audience of 100-200 people gives Meta very little to work with. The resulting lookalike will be inconsistent. Wait until you have 500+ people in your source before relying heavily on lookalikes.

Creating Lookalikes from Low-Quality Sources

A lookalike of all website visitors includes people who bounced after 2 seconds. That is not a quality signal. Use purchase or add-to-cart events for the strongest results.

Never Refreshing Your Lookalikes

Set up dynamic source audiences that update automatically. Static lists get stale as your customer profile evolves and Meta's user base changes.

Overlapping Audiences

If you run a 1% lookalike and a 3% lookalike simultaneously, the 3% includes everyone in the 1%. Use audience exclusions to prevent overlap, or accept that Meta's delivery system will handle some deduplication.

Ignoring Frequency

Watch your frequency metric. When it climbs above 2.0-2.5, your audience is seeing your ads too many times. Either expand to a broader lookalike percentage, refresh your creatives, or both.

Measuring Lookalike Performance

Track these metrics for your lookalike campaigns:

• CPA vs. interest targeting: Lookalikes should deliver equal or lower CPA
• CTR: Should be 1.5%+ for feed placements
• Conversion rate: Compare landing page conversion rates across audience types
• Frequency: Keep below 2.5 for cold prospecting
• ROAS: The ultimate metric, aim for 2x-4x depending on your margins

Key Takeaways

• Lookalike audiences find people who resemble your existing customers using Meta's machine learning
• Purchase-based source audiences produce the best lookalikes followed by add-to-cart and email lists
• Start with a 1% lookalike and expand to broader percentages as you scale
• You need 50+ purchase events before lookalikes become reliably effective
• Exclude existing customers from lookalike campaigns to avoid wasted spend
• Refresh source audiences regularly using dynamic custom audiences that auto-update
• Lookalikes typically outperform interest targeting by 30-60% on cost per acquisition once you have sufficient data