How to Build Lookalike Audiences for Apps (2025 Guide)

Lookalike audiences tell Facebook: "Find me more people like these users."

The quality of that instruction depends entirely on which users you choose as your seed. Point Facebook toward your most valuable users, and lookalikes can deliver comparable CPIs to broad targeting with noticeably better user quality and downstream conversion rates.

Point it toward all users—or worse, all installers—and you're just asking Facebook to find more people who download apps, with no signal about who actually creates value.

Here's how to build lookalike audiences that actually improve performance.

Understanding Lookalike Audiences

Facebook's lookalike modeling analyzes your seed audience and identifies common characteristics—demographics, interests, behaviors, device types, engagement patterns.

It then finds users who share those characteristics but aren't yet connected to your app or business.

Lookalike Percentage Sizing

Lookalikes range from 1% to 10% of a country's population:

1%: The most similar users to your seed. Smallest audience, highest match quality.

5%: Moderately similar users. Larger audience, lower match precision.

10%: Loosely similar users. Largest audience, weakest correlation to seed.

In the US, a 1% lookalike represents roughly 2.3 million people, while a 10% lookalike is 23 million people.

For most app campaigns, 1-5% lookalikes deliver the best performance. Beyond 5%, the similarity becomes so diluted that broad targeting typically performs just as well.

Selecting Seed Audiences

Your seed audience quality determines everything.

Best Seed Audiences for Apps

Purchasers (Best): Users who completed at least one purchase. This is the clearest value signal.

Subscribers: Users on active paid subscriptions. Particularly strong for subscription apps.

High-LTV Users: Top 10-25% of users by lifetime value. Requires LTV calculation but produces the best lookalikes.

Multi-Session Users: Users who opened your app 5+ times. Proxy for engagement when you don't have purchase data yet.

Trial Starters: Users who started free trials. Leading indicator for subscription apps.

Seed Audiences to Avoid

All Users: Too broad. You're telling Facebook to find people who might install any app, not specifically yours.

All Installers: Similar problem. Includes users who installed and immediately churned.

Recent Installers Only: Too small and too new. You don't yet know which of these users will create value.

Minimum Seed Size

Facebook requires 100 users minimum for lookalike creation, but this is far too small to be useful.

Minimum for effectiveness: 500 users

Ideal seed size: 1,000-5,000 users

Maximum useful size: 50,000 users

Beyond 50,000, additional users don't meaningfully improve lookalike quality. Facebook's algorithm identifies patterns from the first several thousand; more data doesn't sharpen the model significantly.

Creating Lookalike Audiences

The technical setup is straightforward, but small decisions impact performance.

Step 1: Create Your Seed Custom Audience

In Facebook Audience Manager:

Click "Create Audience" > "Custom Audience"
Select "Customer List" or "App Activity"
For app activity:
- Choose "People who took specific actions"
- Select your app
- Choose event: Purchase, Subscribe, or your high-value event
- Set timeframe: 30-180 days

Timeframe considerations:

30 days: Most recent users, but smaller audience
90 days: Good balance of recency and volume
180 days: Maximum volume, but may include less relevant older patterns

For apps with steady purchase volume, 90 days typically works best.

Step 2: Generate Lookalike Audience

Click "Create Audience" > "Lookalike Audience"
Select your seed audience
Choose location (start with one country)
Select audience size: 1%
Click "Create Audience"

Multiple Countries:

Create separate lookalikes for each country rather than selecting multiple countries at once. US patterns don't necessarily predict UK or Canadian users well.

Step 3: Test Multiple Sizes

Create 1%, 3%, and 5% lookalikes from the same seed.

Test them sequentially:

Start with 1%
Once 1% spends full budget efficiently, add 3%
Once 3% spends full budget efficiently, add 5%

This prevents spreading budget too thin across untested audiences.

Testing Strategy

Don't assume lookalikes will outperform broad targeting. Test to confirm.

Comparison Framework

Run three ad sets for 7-14 days:

Ad Set 1: Broad targeting (control)

Budget: $15/day

Ad Set 2: 1% Lookalike - Purchasers

Budget: $15/day

Ad Set 3: 3% Lookalike - Purchasers

Budget: $15/day

All other variables (creative, placements, optimization) stay identical.

Success Metrics

After 7-14 days, compare:

CPI: Are lookalikes cheaper, comparable, or more expensive than broad?

Install-to-Purchase Rate: What percentage of installs from each audience complete purchases?

7-Day ROAS: Which audience delivers the best return on spend?

Daily Install Volume: Can the audience support your scale needs?

Common pattern: 1% lookalikes show slightly higher CPI but 30-50% better install-to-purchase rates, resulting in lower cost-per-purchase and higher ROAS.

Advanced Seed Strategies

Beyond basic purchaser seeds, these approaches can improve lookalike performance.

Value-Based Seeds

Instead of all purchasers, segment by purchase value:

High-Value Seed: Users who spent $50+ lifetime

Repeat Purchase Seed: Users with 2+ purchases

These produce lookalikes that more closely match your best users, though you'll need larger total purchase volume to accumulate enough users for the seed (500+ minimum).

Engagement-Based Seeds

For apps without sufficient purchase data:

Power User Seed: Users with 10+ sessions and 5+ days active

Feature Adopter Seed: Users who used your core feature (shared content, created a project, etc.)

These correlate less directly with revenue but work when you need to build lookalikes before accumulating purchase volume.

Exclusion Strategies

Create lookalikes, then exclude existing users:

Create 1% Lookalike - Purchasers
In ad set setup, add exclusion: Custom Audience > App Users > All Users

This prevents showing ads to people who already installed, focusing delivery on net-new users who match your purchaser profile.

Refresh Strategy

Lookalike audiences don't update in real-time as your seed audience changes.

Recommended Refresh Schedule

Seed Audience: Auto-refreshes based on timeframe setting (e.g., "Purchased in last 90 days" updates daily)

Lookalike Audience: Recreate every 30-60 days

Recreating ensures your lookalikes incorporate your newest, most recent high-value users rather than patterns from months ago.

When to Recreate

Monthly for rapidly growing apps adding 500+ purchasers per month
Every 60-90 days for stable apps with slower purchase volume growth
Immediately if you significantly change your product or target market

Lookalikes vs Broad Targeting: When to Use Each

Neither is universally better. Your app's data maturity determines which works best.

Use Broad Targeting When:

You have fewer than 500 high-value users
Your app has mass-market appeal
Previous lookalike tests showed no improvement over broad
You need maximum scale

Use 1% Lookalikes When:

You have 1,000+ purchasers or high-value users
User quality metrics (retention, LTV) vary significantly
You've tested and confirmed better downstream conversion than broad
Slightly higher CPI is acceptable for better user quality

Use 3-5% Lookalikes When:

1% lookalikes are spending full budget efficiently
You need more scale than 1% provides
You're willing to trade some efficiency for volume

Use Both:

Many apps run 70% of budget on broad targeting and 30% on 1% lookalikes, scaling whichever performs better week-to-week.

Common Mistakes

Creating Too Many Lookalikes

Running 10 different lookalike ad sets fragments your budget. Each needs sufficient spend to exit learning phase (10+ conversions within 7 days).

Stick to 2-3 lookalike variations maximum.

Using Lookalikes Without Testing

Lookalikes aren't automatically better. Some apps see worse performance than broad targeting, particularly on iOS where reduced tracking data weakens Facebook's modeling.

Always test before committing significant budget.

Not Segmenting Seeds by Platform

iOS and Android users often have different characteristics and value profiles. Create separate seed audiences by platform for better lookalike precision:

iOS Purchasers > iOS 1% Lookalike
Android Purchasers > Android 1% Lookalike

Ignoring Geographic Differences

Don't create one global lookalike. Build country-specific lookalikes for each market you're targeting:

US Purchasers > US 1% Lookalike
UK Purchasers > UK 1% Lookalike

This accounts for market-specific user characteristics and behaviors.

FAQs

How many users do I need to create a lookalike audience?

Facebook requires a minimum of 100 users, but you need 500-1,000+ users for effective lookalike audiences. Smaller seed audiences don't provide enough data for Facebook to identify meaningful patterns.

What's the best lookalike audience size for app installs?

Start with 1% lookalikes for highest quality. Once they're spending full budget efficiently, test 3% or 5% for additional scale. Audiences larger than 5% typically don't perform better than broad targeting.

Should I use purchasers or all users as my seed audience?

Use your highest-value users (purchasers, subscribers, high-LTV users) as seeds. Lookalikes built from all users typically perform no better than broad targeting since you're telling Facebook to find average users, not valuable ones.

How often should I update my lookalike audiences?

Seed audiences auto-refresh based on your timeframe setting. Recreate the actual lookalike audience every 30-60 days to ensure it incorporates your newest high-value users and current patterns.

Do lookalike audiences work better on iOS or Android?

Android typically shows stronger lookalike performance because Facebook has more complete tracking data to build models from. iOS lookalikes still work but may show less dramatic improvement over broad targeting due to ATT limitations.

Lookalike audiences are only as good as the seed you provide. Focus on quality over quantity—a well-defined seed of 1,000 high-value users will consistently outperform a 10,000-user seed of average installers.