What a customer list can do, and what it needs
Meta's documentation describes customer lists as one source of custom audiencesAn ad audience built from the advertiser's own sources: customer lists, site visitors or engagers.: you upload identifiers for people who already have a relationship with the business, Meta compares them with its own accounts, and the people it matches form the audience. The custom audiences guide sets this source beside website, app and engagement audiences; the difference that matters here is that a list is built from identity rather than behaviour. It reaches customers whose visits the pixel never recorded, because they declined consent, used another device or bought by phone, and it stays exactly as you uploaded it until you replace it.
That makes the first decision the job, because the job decides the segment, the columns and the refresh.
| Job | Segment to upload | Value column | How often to replace |
|---|---|---|---|
| Exclusion from prospecting | Everyone who has bought within the period in which another purchase is unlikely or unwelcome | Not needed | Often enough that recent buyers are always in it |
| Lookalike seed | The customers you want more of, such as repeat or high-value buyers | Yes, where you hold one | When the segment's make-up changes |
| Retention or cross-sell audience | Customers due a replenishment, an upgrade or a related product | Optional | In step with the offer |
Three things decide how useful the result is: whether you may use the data for this purpose at all, how many identifiers each row carries, and whether the list is current. The steps below take them in that order. Where the same customers are also uploaded to Google or LinkedIn, the guide to running customer lists across ad platforms keeps one definition of a customer and one register for all of them.
Before you start
- Access to the ad account in the business portfolio that owns it, with permission to create audiences.
- An export of customers from the shop platform or CRM, ideally from a saved report so the same columns come out each time.
- A decision about the lawful basis for using customer data in advertising, made with whoever owns privacy. Meta's Custom Audience terms require that you have the right to use the data; this page is not legal advice.
- The upload template from the customer list option in Ads Manager, which shows the accepted columns and the formatting each expects.
- A spreadsheet or script that can trim, split and normalise columns without editing the source data by hand.
- The audience register if the account keeps one, so the new list is recorded with its source, purpose and replacement date.
The steps
Decide the job the list will do
Write down which of the three jobs this upload is for, because one file rarely does all three well. An exclusion list wants breadth: every recent customer, including one-off buyers, since the point is to keep all of them out of acquisition spend. A seed wants focus: the lookalike audiences guide explains that a seed of the customers you want more of teaches the model better than a larger, mixed one, so filter to repeat or high-value buyers. A retention audience wants timing. The RFM segmentation framework is a quick way to describe the segment in terms the rest of the business recognises.
Confirm the right to use the data
Before anything leaves the CRM, confirm that these customers' details may be used for advertising on Meta. Meta's Custom Audience terms require that you have that right and, at upload, a declaration of whether the data was collected directly from customers, from partners, or both. HashingConverting an identifier such as an email address into a fixed string with a one-way function, so platforms can match records without the raw value. does not change this. As the guide to keeping personal data out of analytics explains, a hashed email address is still personal data, because anyone holding the address can produce the same hash. Remove people who have asked not to receive marketing, and keep your Suppression ListA list of addresses or contacts excluded from every send, such as unsubscribes, hard bounces and spam complaints. in mind: a person suppressed in email should not reappear in advertising because the two tools hold separate lists.
Export the segment from the CRM or shop
Export only the rows the job needs, with the identifier columns and, for a seed, the value. Use a saved report or query so that next month's export has the same shape. Remove test accounts, staff orders and duplicates of the same person, since a list padded with them distorts both an exclusion and a seed. Keep the export inside the systems that already hold customer data; the file is personal data from the moment it is created.
Choose the identifier columns
Meta documents that email addresses and phone numbers match most reliably, and that adding further identifiers for the same person, such as first name, last name, city, postcode, country and date of birth, improves the chance of a match because Meta can use the combination. Include every identifier you hold and may lawfully use. Expect business lists to match less well than consumer ones: the custom audiences guide notes that people register on Facebook and Instagram with personal addresses, so a file of work emails finds fewer of them.
Clean and format each column to the template
Match the columns and formats the upload template shows, one identifier per column. The faults that most often lower matching are mechanical, and the diagnosis page for an audience that is too small lists them: phone numbers without country codes, first and last names in a single column, stray spaces. Formatting matters even more when you hash the file yourself, because the Conversions API guide notes that an email with capitals or trailing spaces hashes to a different value and matches nothing.
Column Common fault Fix before upload Email Capitals, spaces, several addresses in one cell Trim, lower-case and keep one address per cell Phone Missing country code, punctuation Add the country code and format as the template shows Name First and last name in one field Split into the template's separate columns Postcode and city Mixed formats from old forms Standardise to one format Country Free text Use the code or format the template asks for Add a value column where you hold one
For a seed, add the optional value column. Meta uses it to build value-based audiences and lookalikes that favour people who resemble your higher-value customers, and the lookalike guide recommends margin or lifetime value over first-order revenue where you have it, since that points the model at the customers who make the economics work. Use one consistent measure across the file. An exclusion list does not need values, and leaving them out keeps the file smaller.
Hash the identifiers or let the upload tool do it
Meta documents two routes. You can hash the identifiers yourself, applying the documented algorithm after the documented normalisation rules, or you can upload the formatted file and let the tool hash it in the browser before it is sent. Either way, Meta states that the data is hashed before it leaves your environment and that the uploaded hashes are not kept once matching is complete. Hashing yourself suits teams that automate uploads and want readable data never to leave their own systems; the browser route suits occasional manual uploads. Use one route for the whole file.
Upload, declare the data origin and name the audience
In the Audiences section, create a custom audience from a customer list, upload the file, check that each column is read as the identifier it holds and make the data origin declaration truthfully. Name the audience so that its purpose and date are obvious in every ad set that uses it, for example the job, the segment and the export date, following the account's naming convention. A list called "customers" with no date is how stale exclusions survive for a year.
Read the estimated size and fix a floor label
Once processing finishes, the Audiences table shows an estimated size rather than a precise Match RateShare of the records in an uploaded customer or account list that an ad platform matched to its own users or companies., and small audiences show a floor label instead of a number. Compare the estimate with the rows you uploaded and with your previous upload of the same segment; the match rate entry explains why the figure is read by comparison. If a seed shows the floor label, widen the segment or add identifiers before building anything on it, since Meta documents a minimum seed size for lookalikes on its lookalike page.
Attach the audience as an exclusion, a seed or a retention audience
Use the list for the job you chose. As an exclusion, add it to every prospecting ad set: under Advantage+ audience, included audiences are only suggestions, but excluded custom audiences remain a documented control, which is why the exclusions and overlap guide builds its funnel on them. As a seed, create the lookalike from it. As a retention audience, give it its own ad set and message, excluding recent buyers where a second purchase so soon would be unwelcome. If prospecting still reaches buyers after the exclusion is in place, the page on Advantage+ audience reaching existing customers traces the leak.
Schedule the replacement
A customer list does not update itself. Put the next export and upload in the calendar at the interval the job needs, replace the file rather than appending to it so that people who should leave the audience do, and record the date in the audience register. Check the Audiences table for availability notices at the same time, since the custom audiences guide notes that Meta may flag or remove audiences left unused for a long period.
Special ad categories change what a list can be used for. The custom audiences guide notes that Meta restricts some custom audience and lookalike uses for ads in those categories, enforced when the ad set is created, so check the campaign's category before planning a seed.
A stationery shop's first customer list (illustrative)
Inkmallow Stationery uploads its customers for the first time
Inkmallow Stationery, an invented online stationery shop, has 9,400 customers in its shop platform from three years of trading. Prospecting runs on Advantage+ audience with no exclusions, and the shop's own order data shows 31% of purchases credited to prospecting coming from returning customers. The owner prepares two files from one saved report.
| Row label | Exclusion list | Seed list |
|---|---|---|
| Segment | Everyone who bought in the last 24 months | Customers with three or more orders |
| Rows after removing staff, tests and duplicates | 6,100 | 1,350 |
| Columns | Email, phone, first name, last name, postcode, country | The same, plus lifetime revenue as value |
| First upload, email only | Estimated size shown | Floor label |
| Second upload, all columns cleaned | Larger estimate | Estimated size shown |
The first seed upload carried only email addresses, with capitals and spaces left in, and showed the floor label. After splitting the name column, adding country codes to 1,700 phone numbers and trimming the emails, the seed reported a size large enough to build a value-based lookalike. The exclusion list went on the prospecting ad set the same day. Over the next six weeks, purchases credited to prospecting that came from returning customers fell to 12% in the order data, and the owner set a monthly replacement for the exclusion list and a quarterly one for the seed.
Common mistakes
- One file for every job. An exclusion needs every recent buyer; a seed needs the best ones. Upload each for its purpose.
- Uploading without settling the right to use the data. Hashing is not permission; the declaration at upload is yours to make truthfully.
- Email only. Every extra identifier you may lawfully use gives Meta another way to match the same person.
- Formatting by eye. Capitals, spaces, missing country codes and merged name columns quietly lower matching.
- Leaving out the value column on a seed. Without it, the lookalike learns from whoever buys most often rather than from whoever is worth most.
- Appending instead of replacing. Old files keep lapsed customers in and miss new ones; replace on a schedule.
- Undated names. An audience called "customers" in an exclusion slot hides how stale it is.
