North American Media Experts
Meta lookalike audiences are still one of the highest-ROI targeting tools in paid social. Here's how to build, size, and scale them correctly in 2026.
7 min read
Meta lookalike audiences remain one of the most effective prospecting tools available to paid social advertisers — when built correctly. With iOS privacy changes, third-party cookie deprecation, and Meta's shift toward AI-driven delivery, the way you construct and layer lookalikes in 2026 is meaningfully different from what worked in 2022. This guide covers how to build high-performing lookalike audiences from the right source data, what sizes to use, and how to layer them into a full-funnel campaign structure for Canadian advertisers.
A lookalike audience is a Meta-generated targeting group made up of people who share characteristics with a source audience you define. Meta analyzes hundreds of signals — browsing behaviour, purchase patterns, content engagement, demographics — to find new users likely to behave similarly to your best existing customers.
Lookalikes are prospecting tools. They are not retargeting. They reach people who have never interacted with your brand but statistically resemble those who have. When combined with strong creative and a proper retargeting layer, lookalikes drive the most cost-efficient new customer acquisition available in paid social.
The single biggest mistake advertisers make is building lookalikes from weak source audiences. Garbage in, garbage out — Meta can only find people similar to the seed list you give it.
Source audiences ranked by quality (best to worst):
For Canadian advertisers running paid social campaigns, the minimum viable source audience is typically 100 people. Below that, Meta's modelling becomes unreliable. Aim for 500–5,000 seed users for the strongest signal.
Meta expresses lookalike audience size as a percentage of the country's total population, from 1% to 10%.
Start with 1% and 2% audiences in separate ad sets. Let Meta's delivery system optimize within each. Only scale to 3–5% once you've confirmed performance and have budget to sustain the larger audience without frequency collapse.
A common error is running a single lookalike at 1% and calling it a prospecting strategy. A properly structured paid social funnel uses layered lookalikes tied to your campaign objectives:
Exclude your existing retargeting audiences from each lookalike ad set to prevent overlap. In Meta Ads Manager, add your website custom audience (all visitors, 30 days) as an exclusion from every prospecting campaign.
Meta's Advantage+ campaigns (formerly ASC) have changed how lookalikes fit into campaign architecture. When you run an Advantage+ Shopping Campaign, Meta automatically expands beyond your lookalike targeting if it finds better conversion opportunities elsewhere in its system.
This is not always a bad thing — Meta's AI has access to signals unavailable to manual targeting, including on-platform behavioural data that iOS restrictions don't touch. Many Canadian paid social advertisers are finding that Advantage+ campaigns outperform manually targeted lookalike campaigns by 15–30% on cost per purchase when creative is strong.
The practical recommendation in 2026: run both. Use Advantage+ for your primary conversion objective, and run manual lookalike campaigns alongside to control audience segmentation and creative testing. Feed your best-performing Advantage+ creatives back into manual lookalike campaigns as a test-to-scale mechanism.
With third-party signal loss from iOS 14.5+ still affecting Meta's pixel, the advertisers building the strongest lookalike audiences in Canada are investing in first-party data infrastructure:
Canadian privacy regulations (PIPEDA and provincial equivalents) require that you have consent to upload customer data to Meta's system. Ensure your privacy policy covers this use case and that email lists are sourced from opted-in contacts only.
Lookalike performance is measured against your baseline — either interest-based targeting or broad targeting with no audience constraints. Run an A/B test in Meta Ads Manager with identical creative and budget:
Let it run for 7–14 days with at least 50 conversion events per ad set. In many Canadian accounts, broad targeting now beats 1% lookalikes for purchase campaigns — Meta's algorithm has become skilled enough that removing audience constraints sometimes improves delivery efficiency. If broad wins, use your lookalike as a creative testing environment rather than your primary delivery mechanism.
Benchmark to beat: your 1% purchase lookalike should deliver a CPM within 15–25% of your retargeting campaigns and a cost per purchase within 30–50% above retargeting. If the gap is larger, your source audience likely needs rebuilding with fresher, higher-intent seed data.
Lookalike audiences are not set-and-forget. Rebuild them every 30–60 days if your source is a customer upload (data goes stale), or whenever your business adds 200+ new purchasers or high-value customers to the seed pool. Seasonal businesses in Canada should also rebuild before major demand periods — back-to-school in August, holiday season in October, and Q1 post-holiday for retail.
If your Meta lookalike audiences are underperforming or you're not sure whether your current paid social structure is leaving money on the table, our team can audit your account and build a prospecting architecture tailored to your Canadian market. Contact us for a paid social audit, or request a quote to get started.