How a website custom audience works
Your website sends Meta a record of what visitors do — a page viewed, a product added to a basket, a form submitted — through the Meta pixel in the browser and, where it is set up, the Conversions API on the server. A website custom audience is a rule applied to those records. You choose which event or which page counts, and how far back to look, and Meta assembles the list of people who match.
The retention window is part of the definition rather than an afterthought. A person joins when they match and drops out when their action falls outside the window, so the audience refreshes itself continuously and never needs manual maintenance. Choose a short window and the audience is small but recent; choose a long one and it is larger but full of people whose interest has cooled.
Two limits are worth knowing before you plan around one. The audience cannot be used until enough people are in it, so a newly installed pixel produces nothing usable for a while. And browser restrictions, ad blockers and consent choices mean not every visitor is recorded, which is why a server-side connection matters: it recovers people the browser alone would miss.
Why website custom audiences matter
They are the mechanism behind almost all website retargeting. Someone who read your service page has told you more than any interest category could, and reaching them again with the objection they probably still hold is the cheapest conversation available to most advertisers.
They matter equally as exclusions. Removing recent purchasers from an acquisition campaign stops you paying to sell something people already bought, and removing enquirers stops the same person receiving a first-contact advert while your sales team is already talking to them.
They are also the best seed for a lookalike. A model built from people who reached your checkout describes your business far better than one built from everyone who ever landed on the homepage.
Common mistakes with website custom audiences
Retargeting everyone with the same message is the most common. A person who glanced at a blog post and a person who abandoned a full basket need different adverts, and putting them in one audience wastes the advantage the data gave you.
Windows that are too long are the second. Beyond a certain point you are paying to remind people of a visit they no longer remember. Forgetting exclusions is the third, and the one clients notice: nothing damages trust faster than a customer seeing the offer they just paid for, advertised to them as new. The fourth is planning a retargeting campaign in the first week of a pixel’s life, when the lists are still empty.
How to act on it
Build audiences that match funnel stages rather than one catch-all list: viewers of a specific service page, people who started a form, people who reached the basket, and buyers. Keep a short window and a longer one for each of the important stages so you can compare how quickly interest decays for your product.
Always exclude converters from acquisition campaigns and, where the sale is one-off, from retargeting too. Check the audience sizes every few weeks; a list that stops growing usually means an event has broken rather than that traffic disappeared. And if the numbers look smaller than your analytics reports, that gap is the tracking loss, not a mistake in the audience — closing it is a job for the pixel and server-side setup rather than the audience rules.
