In short
A custom audience is an ad-platform audience built from a business’s own data: hashed customer lists, site visitors, app users, or CRM segments. Every major platform has a version, including Meta Custom Audiences, Google Customer Match, and the equivalents on TikTok and Snap, and they are used for targeting, exclusion, and as the seed for lookalike audiences. Because they are built from data you own, custom audiences are the post-cookie replacement for much of what third-party cookies used to do, which is why they now sit at the centre of most activation strategies. Their quality rests on two things: the coverage and cleanliness of the identifiers you upload, which decides match rate, and how often the audience is refreshed, since a list goes stale as customers change. Uploaded once and forgotten, a custom audience decays; synced from source on a cadence with clean identifiers, it stays an accurate, high-match reflection of who your customers are right now.
The three jobs
Targeting: reach known customers and prospects directly. Exclusion: stop paying to acquire people who already bought. Seeding: give lookalike and similar-audience models a high-quality starting set.
What separates good from bad
Freshness and match rate. A list synced monthly targets churned segments and misses new customers; identifiers uploaded without normalization match a fraction of what they should. Scheduled, hashed, consent-checked syncs fix both.
How do custom audiences get used in a real account?
A retailer builds a lapsed-customer custom audience to win back buyers and excludes recent purchasers to avoid wasting spend. Refreshing both from the CRM weekly keeps the win-back and exclusion lists accurate as customers move between them.
Reference: Meta for Developers, Custom Audiences guide