In 2019, a 1% lookalike audience based on your purchaser list was the most reliable targeting method in digital advertising. Upload your customer emails, let Meta find statistically similar users, and watch CPAs drop. It felt like a cheat code.
Then iOS 14.5 happened. Then iOS 15, 16, 17. Each update eroded Meta’s ability to track user behavior across apps and websites. The signal that made lookalikes powerful — detailed cross-platform behavioral data — got progressively weaker.
Now it’s 2026. Advertisers are split. Some swear lookalikes still work. Others have abandoned them entirely for Advantage+ broad targeting. The truth is more nuanced than either camp admits.
How Lookalike Audiences Work
A lookalike audience starts with a “source audience” — a group of people you define (your customers, your email list, your website visitors). Meta analyzes the characteristics of that source audience and finds other Facebook/Instagram users who share similar traits.
You choose a size (1% to 10%):
- 1% lookalike: The top 1% most similar users in the target country. Smallest but most precise.
- 5% lookalike: Top 5% most similar. Larger reach, less precision.
- 10% lookalike: Top 10%. Broadest. More volume but more diluted.
The quality of a lookalike depends entirely on two things:
- The quality of your source audience data
- Meta’s ability to analyze and find patterns in that data
iOS 14.5+ degraded the second factor significantly.
What iOS Privacy Changes Broke
Before iOS 14.5, Meta tracked users extensively across the internet via the Meta Pixel. It knew which websites you visited, what products you browsed, what you purchased, which emails you opened (via pixel-tracked emails), and how you behaved across hundreds of thousands of apps and websites.
This behavioral data is what made lookalikes powerful. Meta wasn’t just matching demographics — it was matching behavior patterns. “People who browse running shoes on three different sites, click comparison articles, and buy within 48 hours.”
Post-iOS 14.5, approximately 75-85% of iOS users opted out of tracking via App Tracking Transparency. This means Meta lost:
- Cross-app tracking for opted-out iOS users
- Website conversion data for opted-out users (pixel fires but can’t be matched to a Meta profile)
- Behavioral signals that fed the lookalike model
The result: lookalike audiences are built from a smaller, less representative subset of your actual customers. The iOS users who opted out — your customers who use iPhones — are invisible to Meta’s audience modeling.
And iPhone users tend to have higher purchasing power. So you’re building lookalikes from a dataset that systematically excludes a high-value segment.
Do Lookalikes Still Work? It Depends.
When Lookalikes Still Perform Well
Large source audiences (10,000+). The more data Meta has, the better the model works even with reduced signal. If you have 50,000 purchasers in your source audience, the iOS data loss is a smaller percentage of total signal.
Email/phone-based source audiences. Customer lists uploaded directly (hashed emails and phone numbers) aren’t affected by iOS tracking. Meta matches these deterministically to user profiles. This is more reliable than pixel-based source audiences.
High purchase volume. If your pixel still captures hundreds of conversion events per week (because you have high volume and many Android/desktop users), the remaining signal is still meaningful for building lookalikes.
Broad lookalikes (3-5%). Narrow 1% lookalikes are more affected by data loss because the model needs high-quality signal to differentiate the top 1%. Broader lookalikes (3-5%) are more tolerant of signal degradation.
When Lookalikes Underperform
Small source audiences (under 1,000). With limited source data AND iOS signal loss, Meta doesn’t have enough information to build a meaningful model. The lookalike is essentially random targeting.
Pixel-only source audiences. If your source audience is “website visitors” or “people who viewed a product,” iOS opt-outs create a massive gap. Your source audience is missing most of your iPhone visitors.
1% lookalikes in small countries. A 1% lookalike in the US is about 2.4 million people. A 1% lookalike in a country with 5 million Facebook users is 50,000. The smaller the pool, the more noise dominates.
Low conversion volume. If your pixel records fewer than 100 conversions per week, the source data for a purchaser-based lookalike is thin. Each conversion that’s missing due to iOS makes a bigger proportional impact.
Lookalikes vs. Advantage+ Audience
Meta has been aggressively pushing Advantage+ audience (formerly broad targeting or “open targeting”). Instead of you defining the audience, you let Meta’s algorithm find the right people using its aggregate data and your conversion pixel.
How Advantage+ Audience Works
You set:
- Location
- Age (optional)
- Audience suggestion (optional — Meta treats this as a starting signal, not a hard constraint)
Meta handles the rest. The algorithm uses:
- Your pixel conversion data
- Your ad creative (what it shows and who engages)
- User behavior patterns across Meta’s ecosystem
- Real-time auction signals
Head-to-Head Comparison
| Factor | Lookalike | Advantage+ Audience |
|---|---|---|
| Setup complexity | Moderate (need source audience) | Low (set it and go) |
| Data dependency | Needs quality source audience | Needs good pixel + creative |
| iOS resilience | Lower (source audience has gaps) | Higher (uses broader signals) |
| Control | High (you pick the source and %) | Low (Meta decides) |
| Scaling | Limited by source audience size | Scales with budget |
| Learning speed | Faster if source is strong | Slower to start, better long-term |
| Best for | Proven source audiences with 10K+ users | Most campaigns, especially new ones |
The Verdict
For most ecommerce advertisers in 2026, Advantage+ broad targeting outperforms lookalikes because:
- It’s not dependent on a source audience affected by iOS signal loss
- Meta’s algorithm uses real-time signals that are broader than any source audience
- It adapts to creative changes (different ad = different algorithm behavior)
- It scales better (no source audience ceiling)
But lookalikes aren’t dead. They work best as an additional targeting layer, not your only targeting strategy.
The Modern Targeting Strategy
Here’s what high-performing ecommerce advertisers do in 2026:
Primary: Advantage+ Shopping Campaigns
Run your main prospecting budget through Advantage+ Shopping campaigns. These combine Advantage+ targeting with automated creative optimization. They’re Meta’s most algorithmically optimized campaign type.
Secondary: Lookalike Audiences from First-Party Data
Use lookalikes built from first-party data sources that aren’t affected by iOS:
- Customer email list (not pixel-based website visitors)
- High-value purchaser list (customers with LTV above a threshold)
- Repeat purchaser list (customers who bought 2+ times)
- Subscriber list (engaged email subscribers)
Upload these as Custom Audiences and build 3-5% lookalikes. Run them as separate ad sets to test against Advantage+.
Tertiary: Remarketing with First-Party Data
Use Meta’s Conversions API (CAPI) to send server-side conversion data. This recovers some of the signal lost from iOS restrictions. Remarketing audiences built with CAPI data are more complete than pixel-only audiences.
Always: Feed the Algorithm with Good Creative
In a world where targeting precision has decreased, creative quality matters more than ever. The algorithm can only optimize what it has to work with. Give it 3-5 distinct creative angles and let it find the winning combination for each audience segment.
How to Improve Your Lookalike Performance
If you’re committed to using lookalikes, here’s how to maximize their effectiveness in the current landscape:
Use First-Party Source Data
Best sources (in order):
- Customer purchase list (email + phone)
- High-value customer list (top 20% by LTV)
- Repeat purchaser list
- Email subscriber list (engaged segment only)
Avoid using as primary source:
- Website visitors (too affected by iOS)
- Pixel-based custom audiences (incomplete data)
- Video viewers (engagement doesn’t equal purchase intent)
For a comprehensive guide on building a first-party data foundation, see our first-party data strategy guide.
Regularly Update Your Source Audience
Don’t set and forget. Customer behavior and demographics shift over time. Upload fresh customer lists monthly. A lookalike based on last year’s customers may not resemble this year’s ideal buyers.
Test Multiple Sizes
Run 1%, 3%, and 5% lookalikes as separate ad sets in the same campaign. Let Meta allocate budget to the best performer via Campaign Budget Optimization (CBO).
Common results:
- 1% lookalike: Lowest CPA, lowest volume
- 3% lookalike: Moderate CPA, moderate volume
- 5% lookalike: Highest CPA but close to Advantage+
- Advantage+ broad: Often lowest CPA at highest volume (yes, really)
Layer Exclusions
Exclude existing customers from lookalike targeting. You’re paying to find new customers, not re-acquire existing ones. Upload your customer list as an exclusion audience.
Combine with Interest Targeting
“Lookalike AND interested in fitness” is narrower but potentially higher intent than either targeting method alone. Test layered targeting as a separate ad set.
Measuring What Actually Works
The only way to know if lookalikes work for your business is to test properly:
- Run lookalike and Advantage+ ad sets in the same campaign with CBO
- Give each ad set the same creative
- Wait 7-14 days for learning to complete
- Compare CPA, ROAS, and conversion volume
- Check downstream metrics (repeat purchase rate, LTV) — not just first-purchase CPA
Meta’s ad platform will tell you which ad set performs best, but verify with your own analytics. Meta and Google often report different conversion numbers — and Meta’s own ad sets can disagree with each other on attribution.
Your Tracking Determines Your Targeting
Every targeting strategy — lookalike, broad, interest-based — depends on accurate conversion data. If your Meta Pixel and Conversions API aren’t tracking correctly, the algorithm can’t optimize, and your lookalike source audiences are built from incomplete data.
Run a free scan on your site to check your Meta Pixel implementation, verify conversion events are firing correctly, and identify any tracking gaps that are degrading your audience targeting quality.