First Party Data
Facebook Custom Audiences Without Third-Party Cookies: Building Them From Server Events and Customer Data
Facebook custom audiences don't depend on third-party cookies. Meta builds them from four sources: events from your website, customer lists you upload or sync, engagement on Facebook and Instagram, and app activity. Only website audiences depend on your tracking, and what shrinks them isn't the loss of third-party cookies. It's events that never reach Meta, or reach it without anything Meta can match to a person.
This guide covers which audience sources are affected by browser privacy, how events sent through the Conversions API get people into a website custom audience, a worked example of how big your cart abandoner audience really is, and the six audiences we'd build first on a Shopify store.
If you need the Conversions API basics first (what it sends, Pixel vs CAPI, deduplication), read our Meta Conversions API guide. We won't repeat them here.
Do Facebook custom audiences still rely on third-party cookies?
No. The "cookieless" framing causes a lot of confusion, so here is what actually changed.
- Chrome still allows third-party cookies. Google dropped its plan to remove them, and in April 2025 said it would keep its current approach to offering users third-party cookie choice in Chrome.
- Safari and Firefox block them by default. Safari has blocked cross-site cookies by default since March 2020, and Firefox rolled out Total Cookie Protection by default in 2022.
- The Meta Pixel's own identifiers are first-party cookies. Meta's documentation describes
_fbpand_fbcas cookies on your domain holding the browser ID and the click ID.
So the third-party cookie isn't the bottleneck. The losses come from elsewhere:
- Safari caps cookies set by JavaScript, including the Pixel's, at seven days (WebKit, ITP 2.1). A shopper who returns after a week looks new.
- Ad blockers and privacy extensions stop the Pixel script from loading, so no browser event is sent at all.
- In-app browsers keep their own cookie store on iOS, so a shopper who taps an Instagram ad and later buys in Safari looks like two people. Our fbclid guide explains this one.
- Apple's App Tracking Transparency requires apps, Facebook and Instagram included, to ask permission before tracking people across other companies' apps and websites.
Our guide to iOS privacy changes for ecommerce covers Apple's side in more detail.
Which custom audience sources are affected?
| Source | Who gets in | Affected by browser privacy? | Notes |
|---|---|---|---|
| Website | People whose Pixel or Conversions API events match an audience rule and a Meta account | Yes | Retention from 1 to 180 days, per Meta's website custom audience docs |
| Customer list | People in a file or sync whose hashed email, phone or other keys match a Meta account | No | Meta says changes usually take up to 24 hours. No 180-day limit |
| Engagement | People who watched your videos, engaged with your Instagram account or Facebook Page, opened a lead form, and so on | No | Happens on Meta's own apps |
| App activity | People who took actions in your mobile app | Not by browsers, but by app permissions such as ATT | Only relevant if you have an app |
Two things follow from that table. Website audiences are the only ones your tracking setup controls, and they're also the only ones capped at 180 days. Anything you want to target over a longer window, such as everyone who has ever bought, has to come from a customer list.
How do server events get people into a website custom audience?
Meta's documentation is direct about this: "If you share conversion events using Conversions API, you can create a Website Custom Audience and use it for ads." An event sent from your server counts the same way a browser event does.
Three conditions decide whether a given shopper ends up in the audience:
- The event reaches Meta. A server-side event isn't stopped by an ad blocker or a closed tab.
- Meta can match it to a person. Meta's best practices state that unmatched events can't be used for attribution or ad delivery optimization, and an event Meta can't tie to a person can't add that person to an audience either. An event carrying only an IP address and user agent often won't match. One carrying a hashed email, phone,
external_idor thefbcclick ID usually will. - The event fits the audience rule. Event name, time window, and any parameter filters such as URL or value.
The second condition is where most of the difference between setups lies. A server event for an anonymous visitor who never gave an email has little more to match on than the browser event it duplicates. A server event for a shopper you've identified, on this visit or an earlier one, carries the identifiers Meta needs. That's why identity matters more for audience size than the transport.
Running the Pixel and the Conversions API together doesn't double-count people in an audience as long as Meta deduplicates the events. Our pillar covers how Meta deduplicates Pixel and CAPI events.
Worked example: how many cart abandoners does Meta see?
The numbers below are illustrative, picked so the arithmetic is easy to follow. The method is what matters: plug in your own figures from Shopify and Events Manager.
Say 2,000 different shoppers add to cart on your store in a week without buying. The audience Meta builds from them is roughly:
Audience size = shoppers × share whose AddToCart reaches Meta × share Meta can match
| Setup | AddToCart reaches Meta | Meta can match | Cart abandoner audience |
|---|---|---|---|
| Browser Pixel only | 75% (assumed: ad blockers, closed in-app browsers) | 60% (assumed) | ~900 |
| Pixel + server events, no identity | 98% | 60% | ~1,180 |
| Pixel + server events, with known shoppers identified | 98% | 75% (assumed: some carts come from returning customers and subscribers you already know) | ~1,470 |
Read the table from top to bottom. Server events recover the shoppers whose browser event never fired. Identity recovers the ones whose event arrived but couldn't be matched. Neither gets you to 2,000, because some shoppers have no Meta account match and some will never have told you who they are.
To run this on your own store, take AddToCart sessions from your analytics or Shopify's reports, AddToCart events received from Events Manager, and the size estimate Meta shows for the audience once it's built. The gap between each pair tells you which of the two fixes is worth more.
Which custom audiences should a Shopify store build?
These six cover most of what a Shopify brand needs. The last column says which source each one comes from, because that tells you what can go wrong with it.
| Audience | Rule | Use it for | Source |
|---|---|---|---|
| Cart abandoners, 14 days | AddToCart in last 14 days, excluding Purchase in last 14 days | Retargeting | Website |
| Product viewers, 30 days | ViewContent in last 30 days, excluding Purchase | Retargeting with catalog ads | Website |
| Purchasers, 180 days | Purchase in last 180 days | Excluding recent buyers from prospecting | Website |
| All customers | Every customer email and phone, from Shopify or a Klaviyo segment | Excluding past buyers beyond 180 days, lookalike seed | Customer list |
| New customers, 180 days | A new-customer purchase event, if you send one | Lookalike seed built from first-time buyers | Website |
| Video and Instagram engagers | Watched a set share of a video, or engaged with your account | Mid-funnel retargeting | Engagement |
Some notes on these:
- Exclude buyers from both directions. Use the 180-day Purchase audience for recent buyers and the customer list for everyone older. Buyers whose Purchase event never matched only get excluded through the list. Our post on the cost of bad marketing data shows how to price that leak.
- Keep customer lists synced, not uploaded once. Klaviyo's Meta integration syncs a list or segment to a custom audience hourly, and only profiles Meta can match end up in it. Our guide to using Klaviyo data in ads has segment ideas.
- Seed lookalikes from first-time buyers if you want new customers. A seed of all purchasers is weighted toward repeat buyers, who are the easiest people to sell to again and not necessarily the people you're trying to find. Meta needs at least 100 people from one country in a seed.
- Advantage+ campaigns still read your audiences. Even when you don't target custom audiences directly, your existing-customer definitions come from them. Our Advantage+ shopping campaigns guide covers the existing customer budget cap.
How do you build a website custom audience from server events?
Here's the purchaser audience from the table above, step by step.
- Confirm the event arrives from the server. In Events Manager, open your dataset and the Purchase event. Check that it shows both browser and server as connection methods and that the two are being deduplicated.
- Check matching. Look at Event Match Quality for Purchase. If it's low, fix identifiers before you rely on the audience; our Event Match Quality reference explains the scores and parameters.
- Create the audience. In Ads Manager, go to Audiences, create a custom audience and choose Website as the source. Select your Pixel or dataset.
- Set the rule. Choose the Purchase event and a 180-day window. For a cart abandoner audience, add an exclusion for Purchase over the same window.
- Name it after the rule. "Purchase 180d" beats "Buyers." Six months from now, nobody will remember what "Buyers" meant.
- Check the size estimate after a day. If it's far below the number of buyers in that window, go back to step 2.
Custom events work the same way once Meta knows about them. A new custom event shows a warning in Events Manager until someone confirms it; our docs show how to approve Meta custom events.
How Upstack helps with custom audiences
Upstack doesn't upload customer lists to Meta. Klaviyo or a CSV export does that job well. What Upstack changes is the website side: how many events reach Meta, and how many of them Meta can match.
- Server events for every step of the funnel. Upstack Signal sends PageView, ViewContent, AddToCart, InitiateCheckout and Purchase to Meta through the Conversions API, alongside the browser Pixel, with a shared
event_idfor deduplication. - Identifiers on anonymous-looking events. Each event carries the hashed email, phone, name and address Upstack holds for that shopper, an
external_id,fbp, andfbcwhen the visit came from a Meta ad. When a shopper identifies mid-session, Upstack attaches that identity to their earlier events in the session, and the same email or phone links sessions across devices (how stitching works). - Events you can build audiences on. Upstack sends
ncPurchasefor buyers with no earlier order andrcPurchasefor returning buyers, next to the standard Purchase. That gives you the "new customers, 180 days" audience from the table without building anything (setup guide). - Predictive segments. For high-volume Meta advertisers, Upstack Audience scores sessions for purchase intent and sends the highest scorers to Meta through the same Conversions API connection.
Champo, a British haircare brand, identified 128% more customers in 30 days after switching on Upstack's identity resolution, which meant more of its shoppers carried identifiers Meta and Klaviyo could match.
Frequently asked questions
Do Facebook custom audiences need third-party cookies?
No. Website custom audiences are built from events Meta can match to a person, whether they come from the browser Pixel or the Conversions API. The Pixel's own cookies, _fbp and _fbc, are first-party. Customer list and engagement audiences don't involve browser cookies at all.
Can a website custom audience include people tracked only through the Conversions API?
Yes. Meta's website custom audience documentation says that if you share conversion events through the Conversions API, you can create a Website Custom Audience from them. The events still have to match a Meta account to add anyone.
How long can someone stay in a website custom audience?
Between 1 and 180 days, according to Meta's website custom audience docs. Meta removes people automatically when the window passes. For longer windows, use a customer list.
Why is my website custom audience smaller than my traffic?
Two reasons stack up. Some visits never send an event Meta receives, because of ad blockers, closed in-app browsers or a broken Pixel. And some events arrive without identifiers Meta can match. Server-side events fix the first; identifying more shoppers and sending their hashed email, phone and click ID fixes the second.
What's the minimum size for a lookalike seed?
Meta's lookalike documentation asks for a custom audience with at least 100 people from the country you're targeting.
Key takeaways
- Custom audiences come from website events, customer lists, engagement and app activity. Only website audiences depend on your tracking.
- Third-party cookies aren't the bottleneck. Missing events and unmatched events are.
- Events sent through the Conversions API can build website custom audiences, as long as Meta can match them to a person.
- Website audiences stop at 180 days. Use a synced customer list for anything longer, including exclusions of past buyers.
- Measure your own gap: shoppers, events received, and the audience size Meta shows.
Request a demo and we'll compare your website audience sizes with your Shopify traffic.
Want similar results?
Start free for 21 days and see what clean signal does for your performance. No card, no contract.