How an engagement custom audience works
Instead of asking your website who visited, this audience asks Meta’s own platforms who interacted. The sources sit inside Facebook and Instagram: people who followed or visited your Page or professional account, people who messaged you, people who watched a video, people who opened or submitted an instant form, people who responded to an event, and people who interacted with a shop or an Instant Experience.
Each source has its own criteria and its own maximum retention window, and the criteria are more specific than they first appear. A video source, for example, lets you separate people who watched a few seconds from people who watched most of the way through, which is the difference between a thumb that paused and a person who was genuinely interested.
Because the data comes from inside Meta’s platforms, the audience does not depend on your pixel, on browser cookies or on consent banners on your own site. That is the property that makes it useful when website tracking is patchy or when the business has no website at all — a common situation for shops, clinics and restaurants in Nepal that sell through Facebook and WhatsApp.
Why engagement custom audiences matter
They give a warm audience to advertisers who would otherwise have none. If enquiries arrive as Messenger conversations and sales are agreed by phone, the only recorded relationship between a customer and the business may be an interaction on the Page, and this audience is the only way to reach those people again deliberately.
They also make good lookalike seeds where purchase data is thin. A model built from people who submitted a lead form or watched a full video will describe your customer better than one built from a general page-visitor list, even though neither is as good as a list of actual buyers.
Common mistakes with engagement custom audiences
Treating any interaction as interest is the first. A follower from several years ago, or someone who reacted to a giveaway post, has told you very little about whether they want the thing you sell now. Engagement built by boosting entertaining posts is particularly misleading: it fills the audience with people who like the content and have no intention of buying.
The second is choosing the widest source available because it produces the biggest number. Everyone who engaged with the Page is a larger audience than everyone who opened a lead form, and a far weaker one. The third is forgetting the retention limits, so a campaign is planned around a window the platform will not allow. The fourth is failing to exclude people who already converted, which means spending to persuade customers you have already won.
How to act on it
Pick the source closest to a buying decision and work outwards only if the audience is too small to deliver. People who opened or submitted a form, people who messaged you and people who watched a video to the end belong in one group; casual reactors belong in another, addressed with a much softer message.
Keep windows short for the intent-heavy sources and use the longer ones only to build audiences big enough to seed a lookalike. Exclude converters, and exclude anyone already in an active website retargeting audience so the same person is not pursued by two campaigns at once. Then compare cost per result against your website-based audiences rather than assuming either is better.