What Is a Lookalike Audience and When Does It Actually Work?
The audience-building tool that every Meta advertiser uses and most under-configure - and the conditions that determine whether it finds buyers or burns budget.
A lookalike audience is a targeting tool that uses your existing customer data - purchase history, email subscribers, website visitors - to find new people who share similar characteristics. The platform (Meta, Google, TikTok) builds a statistical profile of your best customers and serves your ads to users who match that profile across its user base. When it works, it is one of the most efficient prospecting tools available. When it does not, it spends money finding people who look like your worst customers.
Understanding lookalike audiences is essential for any paid media program targeting cold audiences. Mherie builds lookalike strategy into TTGC's paid programs from day one - but the configuration is almost never what the platform defaults suggest.
How Lookalike Audiences Are Built
You upload a "seed audience" - typically a customer list (email addresses, phone numbers) or a Pixel-based audience (people who purchased, completed a form, or visited a specific page). The platform hashes the PII, matches it to user profiles in its database, analyzes the shared characteristics of that group (demographics, interests, online behavior, purchase patterns), and generates a new audience of users who statistically resemble your seed audience. On Meta, you can set the lookalike percentage from 1% (tightest match, smallest reach) to 10% (broadest match, largest reach). The 1% lookalike is consistently the highest-performing starting point for most accounts.
Seed Audience Quality: The Factor Most Advertisers Get Wrong
The seed audience is everything. Meta recommends a minimum seed of 1,000-2,000 matched profiles for a lookalike to be statistically meaningful - but a list of 5,000 email subscribers who never purchased will generate a lookalike of people who subscribe to things, not people who buy things. For maximum impact: seed from your best customers (highest customer lifetime value), purchasers only (not leads), and recent buyers (last 90-180 days). A 500-person seed of high-LTV customers outperforms a 20,000-person seed of cold email leads every time.
When Lookalike Audiences Work
You have at least 300-500 matched customers in the seed (ideally 1,000+).
The seed is built from revenue events - not engagement events like video views or page follows.
You are running creative specifically designed for cold audiences (problem-aware, not solution-aware).
You have at least $2,000-$5,000/month budget to allow the algorithm sufficient learning events.
When Lookalike Audiences Fail
Your seed is mixed - buyers and non-buyers in the same list - which teaches the algorithm to find people who almost buy.
Your seed is too old - customer patterns from 2022 may not reflect who is buying in 2026.
You are in a very niche B2B category where even a 1% lookalike in a small country is too broad to be useful.
Post-iOS 14 signal degradation has reduced Pixel matching rates below the threshold for reliable modeling.
A lookalike audience is only as good as the customer behavior you teach it to replicate. Garbage in, garbage out - but high-LTV customers in, high-intent prospects out.
TTGC builds paid media programs where lookalike strategy is refreshed quarterly and seeded from revenue events, not engagement signals. Connect your CPA goals to your lookalike structure at a growth assessment.
Build a Lookalike Strategy That Finds Real Buyers
Book a free Brand and Growth Assessment and see exactly how Through The Glass Creatives would approach it.
Sources
- Meta for Business, "Create Lookalike Audiences," Meta Business Help Center, 2025.
- AdEspresso, "Facebook Lookalike Audiences: The Complete Guide," AdEspresso.com, 2025.
- Klientboost, "Lookalike Audience Mistakes Most Advertisers Make," Klientboost.com, 2024.
