Lookalike Audiences: How to Scale Winning Ads

What Are Lookalike Audiences?

A lookalike audience is a group of people who share characteristics with your existing customers. When you tell Meta or TikTok, "Here are the people who bought from me, find more people like them," the platform analyzes hundreds of data points to identify similar users.

Lookalike audiences consistently outperform interest-based targeting by 30-50% once you have sufficient source data. They are the primary scaling tool for dropshipping advertisers.

When to Create Lookalike Audiences

Minimum requirements:

• Meta: At least 100 source events (purchases, add-to-carts, etc.). 500+ is ideal.
• TikTok: At least 100 source events. 1,000+ for best performance.

Do not create lookalikes too early. With only 10-20 purchases, the platform does not have enough data to find meaningful patterns. Use interest targeting until you reach the minimums.

Source Audiences (Seed Data)

The quality of your lookalike depends entirely on the quality of your source audience:

Best Sources (Highest Quality)

1. Purchasers (last 180 days): People who bought from you. The gold standard.
2. High-value purchasers: People who spent above your average order value.
3. Repeat purchasers: People who bought more than once.

Good Sources (Medium Quality)

1. Add-to-cart (last 30 days): High-intent visitors.
2. Checkout initiators (last 30 days): Very high intent, did not complete.

Acceptable Sources (Lower Quality)

1. Website visitors (last 30 days): Broad but less targeted.
2. Video viewers (50%+): Showed interest in your content.
3. Page engagers: Interacted with your Facebook page or Instagram profile.

Rule of thumb: Use the highest-quality source available. A lookalike from 100 purchasers will outperform a lookalike from 1,000 page visitors.

Lookalike Sizes

Lookalike audiences are defined by percentage, representing how closely they match your source:

• 1% Lookalike: Closest match. Smallest audience (roughly 2-3 million in the US). Highest quality.
• 2-3% Lookalike: Broader. Good balance of quality and reach.
• 5% Lookalike: Much broader. Lower match quality but larger reach.
• 10% Lookalike: Very broad. Essentially broad targeting with slight optimization.

Which Size to Use

• Start with 1% for highest performance
• Test 1% vs 3% vs 5% to find the optimal balance for your product
• Use larger percentages when scaling and your smaller audiences are saturated
• Layer with interest targeting for a hybrid approach (1% lookalike + fitness interest)

Campaign Structure with Lookalikes

Testing Phase

Campaign: Sales - Lookalike Testing
  Ad Set 1: 1% Purchase Lookalike
  Ad Set 2: 1% Add-to-Cart Lookalike
  Ad Set 3: Interest Targeting (control)

Scaling Phase

Campaign: Sales - Scaling
  Ad Set 1: 1% Purchase Lookalike (proven winner)
  Ad Set 2: 3% Purchase Lookalike (expansion)
  Ad Set 3: 1% ATC Lookalike (secondary)
  Ad Set 4: Retargeting (warm audiences)

Refreshing Lookalike Audiences

Lookalike audiences should be refreshed regularly:

• Monthly: Update the source audience with recent customer data
• Quarterly: Create entirely new lookalikes from updated source lists
• When performance declines: Refresh the audience as a first troubleshooting step

Lookalike Audiences on TikTok

TikTok's lookalike system is similar to Meta's with some differences:

• Source audience minimum: 100 events
• Size options: Narrow, Balanced, Broad (instead of percentages)
• Start with Narrow for highest quality
• TikTok's algorithm is more discovery-oriented, so lookalikes can be effective with smaller seed sizes

Common Mistakes

• Creating lookalikes with too little source data. Wait until you have 100+ events minimum.
• Only testing 1% lookalikes. Sometimes 3% or 5% outperform 1% for specific products.
• Never refreshing. Stale lookalikes degrade in performance over time.
• Excluding the source audience. Always exclude your source audience from the lookalike campaign to avoid overlap.
• Ignoring interest targeting. Lookalikes are not always better. Keep testing interests as a control group.

Key Takeaways

• Lookalike audiences outperform interest targeting by 30-50% once you have sufficient data
• Wait until you have 100+ purchase events before creating purchaser lookalikes
• Start with 1% lookalikes and test larger sizes for scaling
• Use the highest-quality source available with purchasers being the gold standard
• Refresh lookalikes monthly to maintain performance
• Always keep an interest-based ad set running as a control and backup