Facebook Ads Lookalike Audiences: Complete Guide to Scaling in 2026

You have a webshop running smoothly. Your Facebook Ads are delivering solid results to the Custom Audiences you have already built. But now you want to scale — reach new people who do not yet know your brand, but who are highly likely to be interested in what you sell. That is exactly where Lookalike audiences come in.

The problem is that many advertisers use Lookalikes incorrectly. They pick a weak source audience, set it to 1% and let it run on a small budget. Or they do not have their data foundation in order — the pixel is tracking incorrectly, iOS has limited visibility, and the Lookalike algorithm has too little to work with.

In this guide we cover everything you need to know about Lookalike audiences in 2026: what they are, which source audiences perform best, how to choose the right percentage size, and when to use Advantage+ Audience instead. We are a Facebook Ads agency in Aarhus that works with Lookalikes daily for advertisers across Denmark, and the advice below is based on real campaign data — not theory.

1. What is a Lookalike Audience?

A Lookalike Audience is an audience that Meta builds by analysing your existing customer database — called a "source audience" — and then finding new users who resemble them as closely as possible. Meta uses machine learning to identify hundreds of signals: demographics, interests, behavioural patterns, purchase history, device type, scroll behaviour and much more.

The principle is simple: if your best customers are women between 28–42 in larger cities who are interested in interior design, shop online multiple times a month and typically use an iPhone — Meta finds all other users who match that pattern but do not yet know your brand.

It is important to understand that Lookalikes are not just "broad interest targeting with extra steps." The algorithm works with far more granular signals than the interest categories you can manually select in Ads Manager. That is why Lookalikes have historically delivered significantly lower CPA than interest-based ad sets — because the data foundation is stronger.

2. Source audiences: Custom Audience basics

Before you can create a Lookalike, you need a source audience. In Meta's ecosystem this is a Custom Audience — a list of users you have already interacted with. There are four main types:

Pixel-based Custom Audience: Built from data collected via the Meta Pixel and Conversions API (CAPI). This includes visitors to specific pages, users who have completed certain events (purchase, add to cart, lead), and users segmented by time (e.g. the last 30 days vs. 180 days). It is typically the strongest data source — but requires proper tracking setup.

Customer list (CSV upload): You upload a CSV file with emails, phone numbers or other identifiers, and Meta matches them against Facebook profiles. Match rates in Denmark typically sit at 50–70% depending on data quality. The advantage is that you have full control over who is in the list — you can segment by order value, number of purchases, or other CRM data that Meta does not have access to.

Engagement Custom Audience: Users who have interacted with your content on Facebook or Instagram — watched a video, filled out a Lead Ad, engaged with your page, opened an Instant Experience. This is a good source for top-of-funnel Lookalikes, but the signal is weaker than purchase data.

App Activity: For companies with a mobile app — users who have installed, opened or completed specific actions in the app.

3. The 5 best source audiences for Lookalikes

Not all source audiences are created equal. Here are the five that consistently deliver the strongest Lookalike results — ranked by signal strength.

One important principle: always test multiple source audiences against each other. We have seen cases where an Add to Cart Lookalike outperformed a Purchase Lookalike — simply because the source volume was higher and the signals more stable. Let data decide, not assumptions.

4. Lookalike % — 1% vs. 5% vs. 10%

When you create a Lookalike, you choose a percentage size from 1% to 10%. The percentage indicates what share of the population in the selected country is included — 1% contains the users who most closely resemble your source audience, 10% includes users who resemble it broadly.

In Denmark (approximately 4.8 million Facebook users) the sizes look roughly like this:

A classic mistake is to start with 1% and assume it is always best because it is most "precise." In reality, 1% in Denmark can be too small for Meta's algorithm to optimise effectively — especially with daily budgets above DKK 500. The targeting pool becomes exhausted too quickly, CPM rises, and frequency becomes too high.

5. Lookalike stacking and test structures

One of the most effective ways to scale with Lookalikes is to test several against each other in the same campaign. Here is the structure we use at Gezar for most advertisers:

CBO structure with 3–5 Lookalike ad sets

Campaign Budget Optimization (CBO) lets Meta's algorithm distribute the budget dynamically to the ad sets that perform best. Combined with multiple Lookalike variants, the setup typically looks like this:

• Ad set 1: 1% Lookalike — Purchase (last 180 days)
• Ad set 2: 2–3% Lookalike — Purchase (last 180 days)
• Ad set 3: 1% Lookalike — Top 25% LTV customers (CSV upload)
• Ad set 4: 2–3% Lookalike — Add to Cart (last 90 days)
• Ad set 5: 5% Lookalike — Purchase (scaling test bed)

All ad sets run with the same ads, so the only variable is the audience. Set the campaign budget so each ad set has at least DKK 200–300/day on average, and give the CBO a minimum of 5–7 days to learn before evaluating. Meta's algorithm needs approximately 50 conversion events per ad set per week to exit the learning phase.

Lookalike stacking (advanced)

A more aggressive method is "Lookalike stacking" — where you combine multiple Lookalike audiences in a single ad set. For example a 1% Purchase Lookalike + 1% LTV Lookalike + 1% Add to Cart Lookalike in one combined ad set. This gives Meta a broader data foundation to optimise against, without needing a large budget spread across many ad sets. It works particularly well for advertisers with limited budgets.

6. Advantage+ Audience vs. Lookalike

In 2025–2026 Meta has pushed hard on Advantage+ Audience — their AI-driven targeting solution that in principle replaces the need for manual Lookalikes. The question is: should you drop Lookalikes entirely?

Short answer: no. But you should test both.

Advantage+ Audience uses all available signals — your pixel data, conversion history, ad content and Meta's own user data — to automatically find the users most likely to convert. You can provide "audience suggestions" (previous targeting inputs) as a starting point, but Meta's AI can go beyond them if it finds better opportunities.

When Advantage+ wins

• You have high conversion volume (100+ events/week) and robust pixel tracking
• You are running broad prospecting to cold audiences and want to let the algorithm find the best users
• You lack the time to build and maintain multiple Custom Audiences and Lookalike variants
• Your source audience is too small (<1,000 people) to build stable Lookalikes

When Lookalikes still win

• You have strong first-party data (customer lists, segmented CRM data) that Meta does not have access to
• You want precise control over which signals the algorithm optimises against (e.g. only top customers, not all customers)
• You are in a niche with low volume where Advantage+ lacks enough data to optimise
• You want to A/B test specific source audiences against each other to understand what drives performance

7. When Lookalikes do NOT work

Lookalikes are not a magic solution. There are situations where they do not deliver — and it is important to recognise these early so you do not waste budget.

• Source audience too small (<500 people): Meta's algorithm does not have enough data points to identify meaningful patterns. The result is a Lookalike that resembles a broad, unspecific audience. Build your source first.
• Wrong signals in your source: If your source audience includes returns, complaints or low-quality users, Meta learns to find more of them. Segment your source by quality, not just quantity.
• iOS ATT has decimated your pixel data: If the majority of your traffic is iOS users and you only use the browser pixel (without CAPI), Meta may only see 40–60% of your actual conversions. The result is a Lookalike built on incomplete data.
• You are tracking the wrong events: A Lookalike based on PageView or ViewContent is almost worthless — the signal is too broad. Use events that reflect genuine intent: Purchase, Lead, Add to Cart.
• Audience fatigue in a small market: Denmark has only ~4.8 million Facebook users. A 1% Lookalike of ~48,000 people can be exhausted relatively quickly with daily budgets above DKK 500. When frequency exceeds 3–4, performance drops sharply.
• Poor creatives: The most precise Lookalike in the world cannot compensate for ads that fail to grab attention. If your creatives are weak, the audience is not the problem — the ad is.

8. Budget tips and scaling rules

Scaling with Lookalikes is about finding the right balance between budget, audience size and frequency. Here are the rules we use in practice.

Practical scaling rules

• Start small, scale gradually: Increase your daily budget by a maximum of 20% at a time and wait 3–5 days between increases. Large budget jumps (e.g. from DKK 300 to DKK 1,000) push the campaign back into the learning phase.
• Monitor frequency daily: When frequency on a Lookalike ad set exceeds 3 within a week, it is time to either widen the percentage (e.g. from 1% to 3%) or add a new ad set with a broader Lookalike.
• Horizontal scaling over vertical: Instead of throwing more budget at a single ad set, add new ad sets with different source audiences or percentage sizes. This gives Meta more options and reduces the risk of audience fatigue.
• Use budget minimums correctly: Set ad set minimum spend in CBO campaigns to 70–80% of your average budget per ad set. This ensures all ad sets receive enough budget to learn, while still giving the CBO the freedom to optimise.
• Refresh your source audiences quarterly: Even dynamic pixel audiences should be reviewed. Remove old data (>180 days), update CSV lists, and build new Lookalikes based on updated customer data.

Want to know what a professional Facebook Ads setup costs? We have a complete pricing guide covering everything from management fees to recommended ad spend levels.