Interest-based targeting on Meta - "people who like interior design and art, aged 28-55, female" - has its uses. But it's a broad approximation. You're reaching people who have expressed some interest in a category, not people who actually look like the people who buy from you.
Lookalike Audiences are different. You give Meta a list of your actual buyers (or video viewers, or engaged followers), and Meta finds millions of people who share similar characteristics - demographics, behaviours, browsing patterns, purchase history across the platform. The resulting audience is, in theory, more likely to respond to your ads than a generic interest audience.
In practice, a well-built Lookalike Audience consistently outperforms interest targeting in Artvertise client accounts. This guide explains how to build them correctly and how to use them.
How Lookalikes work
When you create a Lookalike Audience, you give Meta two things: a source audience (your Custom Audience) and a percentage (1-10%).
The percentage refers to the proportion of a country's active Meta users that Meta will match to your source. A 1% Lookalike of the UK contains roughly 420,000 people. A 10% Lookalike of the UK contains roughly 4.2 million.
The smaller the percentage, the more similar the audience is to your source. A 1% Lookalike is tightly matched. A 10% Lookalike is a loose approximation. For most art stores, 1-5% is the useful range.
The quality ranking of source audiences
Not all source audiences produce equally useful Lookalikes. The quality of the output depends on the quality of the input. Here's how to think about source audience quality, from best to worst:
1. Purchasers (best)
This is the best possible source for a Lookalike. Your buyers are the people most aligned with what you're selling - they liked the work enough to pay for it. A Lookalike built from purchasers finds people who share characteristics with your actual customers, not just people who looked at a page.
To use this as a source, you need at least 100 people in the audience for Meta to work with. More is better - 200-300+ gives Meta more data to find patterns. If you're early stage and have fewer than 100 buyers, use a longer time window (180 days) or combine with your high-engagement audience.
2. High-value customers
If you can export your top 25% of customers by lifetime spend from Shopify or your email platform, this makes an even stronger source than all purchasers. Meta will find people who look like your highest spenders - not just anyone who bought once.
This requires a bit of data work (exporting and segmenting your customer list) but for artists with a large enough customer base, it's worth doing.
3. Add-to-cart
People who added items to cart but didn't purchase. This is a warm but not converted audience - they showed strong intent. A Lookalike built from this group performs well when you don't have enough purchasers for a reliable source.
4. Website visitors
All traffic to your site. This is a much noisier source - it includes people who bounced in 3 seconds, people who came from a search for something entirely different, and people who were just browsing. The Lookalike built from this group tends to be broader and less targeted.
Useful when you don't have enough purchasers, but step down to this source, not up to it.
5. Instagram and Facebook engagers
Profile visitors, video viewers, post engagers. This is a decent source if you have a large, genuinely engaged following - people who are consistently watching your process videos or saving your posts. If your engagement is mostly from existing fans and fellow artists, though, a Lookalike built from this group will find more artists, not more buyers.
Use this source when you have a large, actively engaged following and limited purchase data.
How to create a Lookalike Audience
- In Meta Business Manager, go to Audiences
- Click "Create Audience" and select "Lookalike Audience"
- Select your source audience (your purchaser Custom Audience, for example)
- Select the target country or countries
- Set the audience size percentage (start with 1%)
- Create the audience - it takes 1-6 hours to populate
Create separate Lookalikes for different percentages: 1%, 2%, and a 3-5% range. These become separate ad sets you can test against each other.
Size considerations
1% Lookalike: Most similar to your source. Best for performance-focused prospecting. Smaller reach but higher quality.
2-5% Lookalike: Broader. Good for scaling spend when a 1% Lookalike is working and you need more volume.
10% Lookalike: Very broad. Closer to an interest-based audience in character. Only useful for very large-scale awareness campaigns with big budgets.
For most Artvertise clients, the 1% Lookalike is the primary prospecting audience. Once that's consistently profitable, scaling into 2% is the natural next step.
Country considerations
Where you build your Lookalike matters. A Lookalike built in the UK will find UK Meta users. A US Lookalike will find US Meta users.
If you sell internationally, build separate Lookalikes for each major market. Don't build a combined UK+US Lookalike - Meta will weight the audience toward whichever country has more data, which may not be what you want.
For most independent UK artists: start with a UK Lookalike from your purchasers. Once that's working, build a US version (the US market is much larger and often performs differently).
How to test Lookalikes
Run your Lookalike audiences as separate ad sets with the same creative and budget, and compare results over 7-14 days. Three ad sets to compare initially:
- 1% Lookalike from purchasers
- 2% Lookalike from purchasers
- 3-5% Lookalike from purchasers
Look at cost per purchase (or cost per link click if your pixel data is limited) across the three. In most cases, the 1% will have the lowest CPA and the 3-5% will have higher reach but higher CPA. Once you see which is most efficient, allocate more budget there.
Also test Lookalikes against your interest-based audiences. In most mature accounts, Lookalikes win. But in early-stage accounts with small source audiences, interest targeting can sometimes outperform a thin Lookalike - because a Lookalike built from 50 purchasers is less reliable than one built from 500.
When Lookalikes underperform
Source audience is too small. If your purchaser list has fewer than 50-100 people, Meta doesn't have enough data to find meaningful patterns. In this case, use website visitors or Instagram engagers as a temporary source, and continue building your buyer audience until you have enough data.
Source audience is the wrong one. A Lookalike built from all website visitors will include too much noise. Make sure your source is as high-quality as possible.
Poor Pixel data. If your Pixel isn't properly tracking purchase events - because it's installed incorrectly, there's a cookie consent issue, or the purchase event isn't firing - Meta is working with incomplete data. Check your Pixel events in Events Manager before building purchase-based Lookalikes.
Market is saturated. In very small markets (small countries, very niche audiences), a 1% Lookalike might be too small to deliver effectively. Consider broader percentage ranges or expanding to new markets.
If you want to know whether your Lookalike Audiences are built correctly and how to use them in your current campaigns, Artvertise offers a free ad account audit. We'll review your targeting setup and tell you exactly what to adjust. [Book your free audit here.]
Want this done for you?
Get a free audit →