What is Customer Match?

Customer Match moves your own customer list into Google's advertising environment. Instead of building the audience from browsing behaviour, the advertiser starts from data it already holds: buyer email addresses, subscriber phone numbers, postal addresses from a loyalty programme. Those records turn into a segment you can use in Search, the Shopping tab, YouTube, Gmail and Display.

The difference from classic remarketing lies in where the signal comes from. Remarketing needs an identifier to survive in the browser; here the condition is that the person has a Google account and is signed in to it while the ad is shown. That also shifts responsibility, because the data leaves the advertiser's system rather than the user's device.

In Spanish the concept is rendered as segmentación por datos propios, while the product name stays English in the interfaces and in the documentation. The term is worth settling early, because in client conversations it gets confused with bought mailing lists, which is exactly what the feature does not allow.

The process starts on the advertiser's side. The accepted fields are email address, phone number, first name, last name, country and postal code, plus the mobile device ID as a standalone field. Before uploading they have to be normalised: strip leading and trailing whitespace, lowercase the text, write phone numbers in E.164 format, and remove the periods before the domain name in gmail.com and googlemail.com addresses.

The personal fields are hashed with the SHA256 algorithm before they leave. The advertiser can apply the hash itself or let Google Ads do it during the upload; country and postal code are not hashed. Google compares the hashed values against the hashed values associated with Google accounts and, where they coincide, adds that account to the segment. For postal addresses the matching runs on keys built from the hashed name plus the address data. Records that find no partner may be used in policy compliance checks, but they feed neither Customer Match nor any other product, and the file is marked for deletion once the process finishes.

Hence the familiar surprise: the resulting audience is always smaller than the list you uploaded. There are addresses with typos, phone numbers without a country code, people who bought with an address they do not use on their Google account, and people with no account at all. None of those cases is visible one by one, because the interface never returns which specific record found a partner. Only the aggregate size of the segment stays visible. To remain eligible, a list needs at least one hundred members added or updated in the last 540 days, and membership expires after 540 days.

The operational value shows up quickly. You can talk to people who already bought, exclude people who already bought so you do not pay twice for the same customer, and ask Google for similar audiences built from the segment. That last option moves the most budget in large accounts, because it widens reach without depending on cross-site tracking.

The delicate part is permission. Google only accepts information collected in a first-party context, meaning what the customer handed to the advertiser directly on its website, in its app, in its physical shop or in an equivalent situation. Bought or rented lists and data passed on by third parties fall outside, as does information from children under thirteen or from services directed at children. The advertiser must also state in its privacy policy that it shares customer data with third parties to provide services, and obtain consent where the law or Google policies require it. The policy further forbids using data from these campaigns to infer sensitive interest categories about customers.

Uploading a bought database is not a shortcut with manageable risk. It breaches the policy and exposes the account to loss of access or suspension, quite apart from whatever the applicable data protection law says in each country, which we do not assess here.

• Normalise and validate the file before uploading: lowercase, no stray whitespace, phone numbers in E.164 format and country filled in.
• Upload several matching keys per person when you have them, for example email address and phone number, so one mistyped value does not sink the whole record.
• Document where each record came from and keep evidence of how it was collected, so you can answer a compliance review without reconstructing the story afterwards.
• Review the privacy policy before the first upload and check that it mentions passing data to service providers for the provision of the service.
• Refresh the lists on a fixed cadence: membership expires after 540 days, and a list without one hundred members updated in that window stops being eligible.
• Work the exclusions as hard as the targeting. Removing recent customers from an acquisition campaign often improves cost per acquisition faster than raising the bid.

• Uploading a bought or rented database, or data passed on by a third party, trusting that the supplier already obtained the necessary permission.
• Reading the size of the segment as a technical fault and uploading the same file again expecting a different result.
• Using the segment to infer sensitive categories about customers, which the policy expressly forbids.
• Forgetting the 540-day expiry and discovering months later that the audience stopped serving ads.
• Treating hashing as if it were anonymisation and skipping the legal basis because of it. The hash protects the transport, it does not replace permission.