Custom audience identifiers are the pieces of customer data Meta uses to decide whether a person belongs to one of your audiences. On a customer list they are the values you upload, such as email, phone, name and country, which Meta hashes and matches against profiles. On a website audience they are the values the Meta Pixel and the Conversions API attach to events, such as the browser ID, the click ID and the email from a form.
Two kinds of identifier, two different mechanics
A customer list starts with you. You export contacts from a CRM or a spreadsheet, and Meta hashes the values before matching them, so the clear text never sits on its side and cannot be recovered from the hash. What comes back is a match: this person is a Meta account, or is not. Meta says it learns no new identifying information from the hashed data, which is why a list cannot be used to enrich what you already hold.
A website audience works the other way round. There is nothing to upload. Meta listens to the events your pixel and server send, applies the rule you wrote, such as a visit to a URL or a Lead event, and adds the people behind those events for the retention period you chose. Website audiences also need an active Meta Pixel to build at all.
Engagement audiences are the third shape. They are built from people who watched a video, opened a lead form or interacted with a Page or an Instagram account, and they run on a rolling window: with 30 days set, someone who engaged 29 days ago is inside today and outside tomorrow unless they engage again.
The identifiers, in the order they help
| Identifier | Where it comes from | Notes |
|---|---|---|
| the form, or your CRM export | the identifier most lists are built on | |
| Phone | the form, or the export | send it with the country code |
| First and last name | the form, or the export | adds signal when spellings match |
| Country, city, postcode | the export, sometimes the form | useful on large lists |
| Date of birth | the export | rarely asked on a lead form |
| External ID | your own reference for the lead | keeps your records and Meta’s events aligned |
Browser ID (fbp) | the cookie the Meta Pixel sets | website audiences only |
Click ID (fbc) | the _fbc cookie or the fbclid in the URL | ties the audience to an ad click |
| Lead ID | Instant Forms | the 15 to 17 digit ID of a form submission |
| IP address and user agent | the request that submits a form | never uploaded in a list |
Two practical rules. Phone numbers carry the country prefix even when every contact is domestic, because a number without it hashes into a different string. And the browser and click IDs belong to events, not to files: a customer list has no way to supply them, which is why a list and a website audience of the same people are not the same audience.
What raises and lowers the match rate
Audience Manager reports the result as a match score out of ten. The shape of the list decides most of it:
- More identifiers per record is better than more records with one field. An email alone is a thin match; email plus phone plus country and postcode is a strong one.
- Correct normalisation decides whether the hash can match at all. Spaces, capital letters and a missing country prefix each produce a different hash for the same person.
- Freshness matters. A list goes Outdated after 30 days without an update, and the underlying data decays as people change numbers and addresses.
- Size has a floor. A small list cannot absorb much spend, and a match score on a few hundred rows says little.
On the website side, the equivalent lever is the event payload. A Lead event carrying an email, a phone, the click ID and the browser ID matches far better than a PageView that carries none of them, and the audience built on the Lead event reflects that difference. The scoring is explained in Event Match Quality for lead generation.
Where identifiers create a problem
Meta flags audiences that include information its terms do not allow, naming a specific health condition or a credit score as examples. A flagged audience cannot be used in new campaigns, and campaigns already using one should be edited or paused, which usually means removing the data, building the audience again or requesting a review.
The data source behind an audience can be restricted too. Meta may place a source in a category such as health and wellness or financial service, and some categories bring core setup, which strips custom parameters and everything in a URL after the domain. An audience whose rule depends on a parameter, or on a path such as a thank-you page, then fills slowly or stops updating. The place those categories live is described in Meta dataset.
One structural limit is worth knowing before you plan a retargeting ladder: a website audience fills forward and Meta does not backfill, so the first day is empty by design and the first week is thin. Meta also warns it may miss events from some iOS 14 users, which bites harder on a small site than on a busy one.
Using identifiers well in lead generation
- Decide the job of the audience: retarget people who opened a form without submitting, or exclude the leads you already own from prospecting.
- Build the source data first. A clean customer list of past leads, with email and phone, is the most useful asset in a lead generation account, mostly as an exclusion.
- Normalise before you upload: lowercase, trimmed, country prefix on phones, two-letter country codes.
- Upload a full file rather than a partial one, and refresh it on a schedule shorter than 30 days.
- Read the match score after a day, and add a field that the list is missing rather than adding more rows.
- Attach retargeting audiences as hard limits, not suggestions, because a custom audience added under suggestions is one Meta may deliver beyond.
- Keep an eye on size against budget. A list too small for the daily spend throttles delivery and looks like a creative problem.
The exclusion in step 2 is the one that pays first: uploading the leads you already have and excluding them stops you paying to collect the same person twice. Excluding an audience does not switch Advantage+ off. And for a lookalike, the same identifiers decide the seed quality, which is covered in lookalike audience.
The mechanics of each audience type, including the 500-audience limit per ad account and the five rules per website audience, are in custom audience. When one refuses to fill, the order to diagnose it in is in why is my Meta custom audience not working?. And because an audience only changes who sees the ad, whether it lowered the cost per lead shows up in the leads themselves: DropCPL counts each one on your landing page and keeps it tied to the ad, the creative and the landing page, with spend synced per ad from the Meta Ads API.