How a lookalike audience actually works
You give the platform a source audience, a customer list, purchasers, or a high-value segment, and it analyses the traits those people share, then finds a much larger group of new users who resemble them. On Meta you also pick a size, from a tight 1 percent of a country's users that resembles the source most closely, to a broad 10 percent that trades similarity for reach. The mechanism is pattern-matching at scale: the platform cannot know why your customers convert, only what they look like, and it finds more of that look.
Why the source list decides everything
A lookalike inherits the quality of its seed. Build it from all buyers and you get an average customer; build it from your top 10 percent by lifetime value and you get a model aimed at your most profitable customers. Value-based lookalikes, seeded from customers weighted by revenue, consistently outperform ones seeded from raw purchasers, because they teach the platform to find valuable customers rather than merely frequent ones. iClick invests in the source list first, because no audience-size tuning can rescue a lookalike built from a weak or noisy seed.
Lookalikes in the age of automation
As Meta's Advantage+ and Google's automated targeting have matured, explicit lookalikes have become less of a hard boundary and more of a signal. Increasingly the platforms take your source audience as a strong hint about who converts and then explore beyond it, much as audience signals work in Performance Max. That shift makes the seed even more important, not less: when the audience is a suggestion the algorithm expands from, the quality of that suggestion still shapes where the exploration starts and how fast it finds good customers.
The lookalike mistakes that waste budget
Three errors recur. Seeding from a list that is too small or too broad, so the model has no clear pattern to match. Using raw buyers instead of value-weighted customers, which finds cheap, low-value lookalikes. And never refreshing the source, so the audience drifts from who your best customers are today. iClick rebuilds source lists from current, value-ranked customer data on a cadence, because a lookalike is a snapshot of your best customers at one moment, and the business it is trying to grow keeps moving.
