What Is a Lookalike Audience?
A lookalike audience is a powerful prospecting tool available on platforms like Meta Ads and Google Ads. The idea is simple: you already know who your best customers are, so why not find more people just like them? You provide the platform with a seed audience, a group of people who represent your ideal customers, and the platform uses machine learning to identify patterns in that group's attributes. It then scans its entire user base to find new people who closely match those patterns.
The seed audience can come from several sources. You might upload a customer email list, use your website purchasers as recorded by the Meta Pixel or Google tag, or use your email newsletter subscribers. The quality of the seed matters enormously. A seed built from your highest-value purchasers will produce a much more accurate and responsive lookalike than one built from all website visitors, many of whom may have bounced immediately without any meaningful engagement.
On Meta, you control how closely the lookalike mirrors your seed by setting a percentage. A 1% lookalike gives you the smallest and most similar audience. Expanding to 5% or 10% increases reach but loosens the similarity. For South African campaigns, the total Facebook and Instagram audience is substantially smaller than in markets like the United States or United Kingdom, so finding the right balance between audience size and similarity quality is crucial. Most South African advertisers find that 1% to 2% lookalikes outperform broader percentages because they maintain the relevance signals that make the audience valuable.
Google's equivalent feature is called Similar Segments, found within Google Ads audience management. It operates on a similar principle, building a prospecting list based on behavioural signals from your existing remarketing audiences. Google automatically suggests similar segments when your source lists meet certain size and activity thresholds, which can vary depending on the industry and market.
Lookalike Audience In Practice
The scenario below is an illustrative example, not a Juicy Designs client result. The figures indicate the scale of effect that lookalike audience work typically produces, so treat them as indicative rather than measured.
Picture a Durban-based insurance broker that has been running Meta Ads for two years and has accumulated a list of around 3,000 paying clients. It would upload this customer list to Meta Business Manager as a custom audience, then create a 1% lookalike audience targeting South Africa. Meta identifies the demographic and psychographic patterns in those 3,000 clients, finds the top 1% of South African Facebook users who match those patterns, and makes this new audience available for prospecting campaigns. Because the audience shares characteristics with people who already converted, it would typically respond better to ads than broad interest or demographic targeting alone.
The combination of lookalike audiences for prospecting with retargeting for conversion creates a complete paid social funnel. Lookalikes fill the top with qualified new prospects. Retargeting catches those who engaged but did not convert. This two-layer approach is the foundation of effective Meta and Google social media advertising for South African businesses seeking to grow without wasting budget on irrelevant audiences.
How lookalike audiences work
A lookalike audience is a group of new people an advertising platform finds because they resemble an existing audience you supply, called the source or seed. You provide a seed, your customers, website visitors, or an engaged audience, and the platform analyses its shared characteristics and behaviours, then builds a larger audience of users who share those traits but whom you have not reached. This lets you extend targeting from people who already know you into fresh prospects likely to be interested, which is how advertisers move from retention into efficient prospecting. As third-party tracking declines, lookalikes built from your own first-party seed audiences have become a central way to find new customers, making the quality of your seed data increasingly important.
Improving lookalike audience quality
The quality of a lookalike audience depends heavily on the quality of its seed, so the most important lever is what you feed it. A seed of your best, highest-value customers produces better prospects than one of all visitors indiscriminately, because the platform models whatever traits the seed shares. The seed also needs to be large enough for the platform to find reliable patterns, and clean, genuinely relevant rather than a mixed bag. The size of the lookalike matters too: a tighter lookalike resembles the seed more closely but reaches fewer people, while a broader one reaches more but resembles the seed less. Testing seed audiences and lookalike sizes, and refreshing seeds as your customer base evolves, keeps the audiences performing.
FAQ
How large does a seed audience need to be for lookalike targeting?
Meta recommends a seed audience of at least 1,000 people for reliable lookalike modelling, with 2,000 to 5,000 producing the best results. Google's Similar Segments require your remarketing list to have sufficient users before activating. In the South African market, building your seed from purchasers rather than all visitors produces a much more accurate lookalike.
What percentage size should a lookalike audience be?
A 1% lookalike is the closest match to your seed audience and typically delivers the best conversion rates. Sizes of 2% to 5% expand reach but reduce accuracy. For South African campaigns where total addressable audience is smaller than in major global markets, starting at 1% to 2% is usually the right balance between reach and relevance.
How large does a seed audience need to be for a lookalike?
Platforms usually need a minimum seed size, often around a thousand or more people, to find reliable patterns, though larger and cleaner seeds produce better results. The seed must also be relevant; a smaller seed of high-value customers can outperform a larger, indiscriminate one.