"Should I use 1% or 10% lookalikes?" It's one of the most common questions in Meta advertising. The answer isn't as simple as picking a number — it depends on your seed audience, budget, objective, and testing strategy.
Understanding Lookalike Percentages
Lookalike percentages represent how closely Meta matches users to your seed audience. A 1% lookalike contains the top 1% of users most similar to your seed. A 10% lookalike casts a wider net, including users with weaker similarity signals.
- 1% Lookalike: ~2.3M people in the US (highest similarity)
- 5% Lookalike: ~11.5M people in the US (moderate similarity)
- 10% Lookalike: ~23M people in the US (broader reach)
The 1% vs 10% Debate
The conventional wisdom says "start with 1% for quality, expand to 10% for scale." But this oversimplifies the reality. Here's what actually matters.
When 1% Lookalikes Win
- Small budgets: Under $100/day, concentrate spend on highest-quality matches
- High-LTV products: When each conversion matters, precision beats reach
- Strong seed audiences: 1,000+ purchasers with clear behavioral patterns
- Niche markets: When your ideal customer is highly specific
When Broader Lookalikes Win
- Large budgets: $500+/day needs larger audiences to spend efficiently
- Low-ticket items: Volume matters more than per-conversion precision
- Weak seed data: Small or low-quality seeds produce unreliable 1% matches
- Broad appeal products: When your product fits many customer profiles
The Seed Audience Factor
Your lookalike is only as good as your seed. A 1% lookalike from 100 random purchasers will underperform a 5% lookalike from 10,000 high-LTV customers. Learn how to build better seeds in our guide to building custom audiences.
Optimal Seed Characteristics
- Size: 1,000-50,000 people (minimum 100, but more is better)
- Quality: High-value customers outperform all customers
- Recency: Recent purchasers (90 days) beat lifetime lists
- Consistency: Similar behaviors within the seed improve matching
Testing Framework
Rather than guessing, test lookalike percentages systematically. Understanding scientific testing methods helps structure these experiments.
Stacked Lookalike Test
Test multiple percentages simultaneously with proper exclusions:
- Ad Set 1: 1% Lookalike
- Ad Set 2: 1-3% Lookalike (excludes 0-1%)
- Ad Set 3: 3-5% Lookalike (excludes 0-3%)
- Ad Set 4: 5-10% Lookalike (excludes 0-5%)
What to Measure
- CPA by percentage: Cost efficiency at each tier
- ROAS by percentage: Revenue quality at each tier
- Delivery stability: Can Meta spend your budget efficiently?
- Frequency: Are you exhausting smaller audiences too quickly?
Budget-Based Recommendations
Your daily budget heavily influences optimal lookalike strategy:
$50-100/day
Stick with 1-3% lookalikes. You don't have enough spend to test broader audiences effectively, and concentration improves learning phase efficiency.
$100-500/day
Test 1%, 3%, and 5% lookalikes. You have room to compare, but 10% may still be too diluted for your budget to optimize properly.
$500+/day
Test full range from 1-10%. At this budget, you need audience scale, and Meta's algorithm has enough data to optimize across broader pools.
Advanced Lookalike Strategies
Value-Based Lookalikes
Use customer LTV data to create value-optimized lookalikes. Meta will find users similar to your highest-value customers, not just any purchasers.
Multi-Seed Lookalikes
Combine multiple seed audiences (purchasers + high-engagement visitors) to create more robust lookalike models with broader signal inputs.
International Lookalikes
For international expansion, 3-5% often outperforms 1% because behavioral patterns vary by market. Your US seed may not translate perfectly to Germany.
How ROASPIG Helps
Lookalike optimization requires continuous testing and iteration. ROASPIG automates the process:
- Seed Scoring: Evaluate seed audience quality before building lookalikes
- Automated Testing: Deploy stacked lookalike tests with proper exclusions
- Performance Tracking: Monitor CPA and ROAS by lookalike tier
- Budget Recommendations: Suggest optimal percentages based on spend levels
- Creative Diversification: Ensure your ads pass the Andromeda diversification test
Common Mistakes
- Using all customers as seed: High-value customers make better seeds
- Ignoring seed size: Small seeds (under 500) produce unreliable lookalikes
- Not excluding purchasers: Always exclude existing customers from prospecting
- Testing without budget: Each ad set needs enough spend for statistical validity
- Copying competitor strategy: Your optimal percentage depends on your unique data
The Bottom Line
There's no universal "best" lookalike percentage. The right choice depends on your seed quality, budget, product, and market. Start with 1-3% if you're unsure, but commit to systematic testing to find your optimal range.
The advertisers who win don't debate 1% vs 10% — they test both and let data decide.
Frequently Asked Questions About Lookalike Audiences
A lookalike audience is a targeting option that finds users similar to your existing customers or website visitors. Meta analyzes your seed audience and identifies users with matching characteristics and behaviors.
Meta requires a minimum of 100 people, but 1,000-50,000 produces the best results. Larger, high-quality seeds give Meta more data points to identify meaningful patterns.
It depends on your budget and goals. 1% is more precise but smaller. 10% offers more scale but less similarity. Most advertisers find their sweet spot between 1-5% for acquisition campaigns.
Lookalikes automatically update as your seed changes. However, recreate them quarterly or when your seed audience composition changes significantly to ensure the model reflects your current best customers.
Yes, but purchaser-based lookalikes typically outperform visitor-based ones. Website visitors include tire-kickers, while purchasers represent validated customer profiles.