Table of content
If you send customer data to Google Ads, whether a list for customer list targeting or offline conversions from your CRM, you now go through Data Manager (Google Ads Data Manager). It is the single entry point for those imports, and the older routes through the Google Ads API were closed during 2026. The action itself has become simple, point and click. The decision behind it has not: the moment you connect a CRM, a data warehouse or a file, you are also deciding which personal data leaves your company, to which country, and on what legal basis.
This article covers the ground in that order: what the tool does, what 2026 changed, what Google actually receives, what the FADP and the GDPR require before the first connection, and why none of this replaces your server-side tracking.
The essentials
What it is: an import and management tool inside Google Ads that connects a data source once and then activates it across several destinations.
What changed in 2026: offline conversion import and enhanced conversions for leads import moved to the Data Manager API and have been blocked in the Google Ads API since 15 June 2026. Google has also announced that enhanced conversions for web and for leads will be merged into a single feature with a single toggle.
What Google receives: contact details normalised and hashed with SHA-256, and nothing more in the Swiss case. A hashed email is not anonymous data. In Switzerland, the EEA and the UK, Google does not support matching by IP address and timestamp, and IP address sharing must be excluded there.
The legal point: a CSV file or a warehouse table carries no consent signal. If your CRM cannot answer, contact by contact, whether advertising personalisation was agreed to, the connection is premature.
What it does not replace: server-side tracking. The two set-ups do two different jobs and are not substitutes.
What exactly is Google Ads Data Manager?
Data Manager is an import and data management tool with a point-and-click interface that brings customer data living outside Google into Google Ads, and then activates it. You find it in Google Ads under the Tools icon. The official help page states its central promise: connect a source once, then use it for several destinations.
Three concepts structure the tool, and they are worth separating straight away, because governance mistakes almost always come from blurring them.
The data source is where the information lives: a linked product, an uploaded file, or a third-party integration such as a CRM.
The connection is the data object actually imported from that source: a table, a file. This is the real scope of what leaves your systems.
The destination is the use case in which the data is activated: customer list targeting, offline conversion import, and at API level, other measurement products.
The Data Manager page also has three areas: connecting first-party data, managing the Google tag, and configuring consent settings. That grouping is not incidental. In the same interface you set both what your site tag sends to Google and what your imports send. It is coherent from Google's side, and a frequent source of confusion on the advertiser's side, because the two flows do not follow the same consent mechanics.
One thing to state plainly: Data Manager is not a customer data platform. It does not store your database for your own use, does not segment for you, does not resolve your identities. It is an intake pipe toward Google's activation products. If you are looking for a customer record system, this is not it.
Why did 2026 make Data Manager unavoidable?
Until 2025, sending offline conversions or a customer list to Google Ads went through the Google Ads API, with dedicated services, or through a manual upload. In 2026 that route closed in stages, in favour of the Data Manager API, released as version 1.0 on 2 April 2025 and generally available with version 1.3 on 6 October 2025. The API release notes date each step, and the documentation presents this API as the platform where future features will be launched exclusively.
Here are the milestones worth keeping, as the official documentation dates them.
6 October 2025, version 1.3: general availability, and the targeting threshold lowered to 100 audience members.
17 February 2026, version 1.5: full audience management, user identifier ingestion, and a resource linking advertisers to data partners.
April 2026: Google Ads accepts data from site tags, Data Manager and API connections at the same time, an overlap period meant to absorb migrations.
7 and 28 May 2026, versions 1.6 and 1.7: store sales conversions for Google Ads, offline conversions to Campaign Manager 360, Search Ads 360 and Display & Video 360, and IP address ingestion for customer list targeting.
15 June 2026: the hard switch. Offline conversion import and enhanced conversions for leads import move to the Data Manager API and are blocked in the Google Ads API.
June 2026, as announced by Google: enhanced conversions for web and enhanced conversions for leads are merged into one feature, turned on or off with a single toggle. Google dates this merger to June 2026 but its help page still writes it in the future tense, so check the actual state in your account rather than assuming it. Once the merger lands, the choice of an implementation method between the Google tag, Google Tag Manager and the Google Ads API disappears.
30 July 2026, version 1.8: removal of all members of an audience in one operation, detailed warnings on ingestion, and new address fields.
The diagram in this section lays out that timeline, from version 1.0 of the API to the 15 June 2026 switch. If you do not have it in front of you, the reading is this: two years of progressive tooling, an overlap period in spring 2026, then a clean shutdown of the old route in June.
Two practical consequences for a Swiss SME or an agency. First, integrations built by a provider who has since moved on may have stopped working in June 2026 with nobody watching: a failing offline conversion import raises no visible alert in campaign reports, it only lowers the volume of conversions coming back. Second, enhanced conversions changed shape: existing accounts that had already accepted Google's customer data terms will be migrated automatically to the new unified status, with no action on their part, according to Google's help page. It is worth checking what state the customer data panel of your accounts is actually in.
Which sources can you connect, and to which destinations?
The list of sources supported for direct connection, as the documentation publishes it, covers databases, warehouses, object stores, files, and a handful of CRM and e-commerce platforms.
Warehouses and databases: BigQuery, Snowflake, Amazon Redshift, PostgreSQL, MySQL, Oracle.
Storage and files: Google Cloud Storage, Amazon S3, Google Drive, Google Sheets, SFTP, HTTPS, with CSV or TSV files.
CRM and platforms: HubSpot, Zoho CRM, Salesforce, ActiveCampaign, Klaviyo, Shopify.
Two exceptions to know before promising anything to a client: Salesforce is supported for offline conversion import but not for customer list targeting, and Shopify is the other way round, customer list targeting only.
At API level the scope is wider than the two destinations visible in the interface. The documentation lists Google Ads, Display & Video 360, Google Analytics and Floodlight for Campaign Manager 360 and Search Ads 360. On data types, it separates audiences, fed by contact details, mobile device identifiers and, for Google Ads, advertiser-assigned user identifiers, from events, which cover offline conversions, multi-source online conversions, store sales, and for Google Analytics multi-source, recommended or custom events.
The diagram in this section shows the full path, from source to destination via the connection. The sequence to remember: one source can feed several connections, one connection can serve several destinations, and it is at connection level, not source level, that the exact field scope of what leaves is decided.
What Google actually receives: hashing, encryption, trusted environment
This is the question clients ask worst, and the one whose answer decides the rest. Today, contact details intended for customer list targeting must be normalised and then hashed with the SHA-256 algorithm before being sent. Google states that it does not receive the email addresses themselves, hashing being a one-way operation. IP addresses and interaction timestamps, which an advertiser may supply to improve match rates, must not be hashed. That point does not apply to Swiss advertisers, however. Google's customer list targeting policy states that matching by IP address and timestamp is not supported in Switzerland, the EEA and the UK, and that IP address sharing must be excluded there. If a provider offers to add IP addresses to your uploads to gain a few points of match rate, the answer is no.
On top of that hashing, Google has added confidential matching. The principle: the overlap between your list and Google's data is computed inside a trusted execution environment built on Google Cloud Confidential Space, and unused identifiers are removed from the imported list. The documentation specifies that confidential matching is on by default and free of charge, with no prerequisites, for customer list targeting when the source is linked through a direct connection. Additional encryption is possible but not required: it adds access control guarantees, including the technical guarantee that only confidential matching can process the data, through the attestation mechanism that lets the environment prove by cryptographic signature which software it runs.
At API level, encryption has a precise, documented mechanic: values are formatted and hashed under the same rules as unencrypted data, the hash is Base64-encoded, then encrypted with a data encryption key using XCHACHA20_POLY1305, with the result encoded in hexadecimal or Base64. Keys are managed with Google Cloud Key Management Service or AWS Key Management Service. Unhashed values such as region code, postal code, city or administrative area are not encrypted.
On retention, Google states that it keeps uploaded files only as long as needed to build the audiences and check policy compliance, and says it does not use them to create or improve customer profiles, nor share them with third parties or other advertisers, subject to legal obligations and binding government requests. The systems supporting customer list targeting are presented as ISO 27001 certified.
One announced change deserves a cautious mention. According to an email sent to advertisers, reported on 18 October 2026 by the trade publication PPC Land and picked up on 22 October by Search Engine Roundtable, Data Manager would stop requiring SHA-256 hashed customer data from 16 November 2026, accepting encrypted data instead, decrypted, normalised and matched inside a hardware-isolated trusted execution environment. As of 2 November 2026, the help and documentation pages consulted for this article still describe hashing as the step before encryption. If you run an import pipeline, this is a point to confirm in the official documentation before changing anything, rather than on the basis of a relayed announcement.
The diagram in this section breaks down what Google receives, step by step. The reading is: normalisation on your side, hashing on your side, optional encryption on your side with your keys, matching on Google's side inside a trusted environment, then an audience or a conversion ready to activate, and nothing more.
One point technical discussions tend to skip over. A hashed email is not anonymous data. It still allows a person to be identified by matching, which is precisely its function. Under art. 5 let. a FADP, which defines personal data as any information relating to an identified or identifiable natural person, a hashed identifier therefore remains personal data, and the whole regime of the law applies to it. Hashing is a useful security measure, not an exit from scope.
What do the FADP and the GDPR require before you connect a source?
Connecting a CRM to Google Ads is both a processing of personal data and a disclosure of data abroad. Three FADP provisions govern the operation directly, and a fourth, contractual question comes on top.
Purpose and proportionality, art. 6 FADP. Processing must be lawful, carried out in good faith and proportionate. Data may only be collected for a purpose that is evident to the data subject, and reuse must stay compatible with that purpose. So the question is concrete: could the person who left an email address to get a quote recognise that this address would be used to build an advertising audience? If the answer is no, the problem is not technical.
The duty to inform, art. 19 FADP. At collection, the person must be informed at minimum of the controller's identity, the purpose, and any recipients. Where data is disclosed abroad, the information must name the country concerned and the applicable safeguards. A privacy policy that does not name Google as a recipient and says nothing about the transfer does not meet this obligation.
Disclosure abroad, art. 16 and 17 FADP. A transfer is possible if the Federal Council has established that the destination country offers an adequate level of protection. On 14 August 2024, the Federal Council added the United States to that list for American companies certified under the Swiss-US Data Privacy Framework, with the amendment to the Data Protection Ordinance entering into force on 15 September 2024. A transfer to a certified company is therefore possible without additional safeguards. Two practical caveats: certification is something you verify, it is not permanent, and it applies to a specific entity, not to a whole country.
Google's own rules. The customer list targeting policy imposes several conditions that read like a specification. You may only upload information customers shared with you directly, or that you collected in a first-party context, on your sites, in your apps or in your physical stores. Your privacy policy must state that you share customer data with third parties acting on your behalf. You must obtain user consent for that sharing where the law requires it, including under Google's EU user consent policy, whose scope covers the European Economic Area, the UK and Switzerland. You may not upload information about users under 13, nor use sensitive interest categories. Finally, a list is valid for 540 days and needs at least 100 members added or refreshed in that window to stay eligible.
That last block is worth underlining for agencies: the fact that Google's consent policy explicitly covers Switzerland settles a recurring debate. A Geneva SME running customer list targeting is in scope, with the attendant obligations on proof of consent and withdrawal, independently of any discussion about the territorial reach of the GDPR over its activity.
Here is the finding that costs the most in practice. A warehouse table or a CSV file carries no consent signal. Unlike a web event, where Consent Mode accompanies each measurement with the consent state, a batch import is silent on the point. The export filter is therefore your only protection, and it has to rest on reliable information in the CRM: a date, a scope, a source. If your CRM cannot answer, contact by contact, the advertising personalisation question, do not connect it. How consent and transmission fit together is covered in our guide to Google Consent Mode v2, and the comparison of the two regimes in our guide to FADP versus GDPR for web agencies.
Does Data Manager replace server-side tracking?
No, and the confusion is common because both set-ups talk about first-party data. They do two distinct jobs.
Server-side tracking measures what happens on your site or app, in real time. It carries the consent state event by event, filters and enriches before sending, and gives you control over what leaves the browser. It is the layer that produces the click identifiers and first-party identifiers everything else rests on. Our complete guide to server-side tracking and FADP compliance details that architecture.
Data Manager handles what happens off the site: a status changing in the CRM, a contract signed three weeks later, an in-store sale, a refund. It sends that information back in batches to the activation products.
The complementarity is easy to state: server-side supplies the identifier, the CRM supplies the outcome, Data Manager carries the outcome back attached to the identifier. Remove any of the three and the chain breaks. That is exactly the mechanism described in our article on offline conversions between CRM and Google Ads, whose governance side this article completes.
The diagram in this section puts the two chains side by side. Read it in three steps: the web chain, from browser to server-side container then to the destinations, in real time and with consent; the offline chain, from CRM to Data Manager then to the same destinations, in batches and with no embedded consent signal; and the junction point, the click or transaction identifier, which must be the same on both sides.
Three traps recur at that junction.
Double counting. If the same conversion goes out client-side, server-side, then a third time from the CRM, it is counted several times. You need discipline around a unique transaction identifier, and a written decision on which source is authoritative for each conversion type.
The import window. An offline conversion has to arrive within the window the conversion action accepts. A monthly export sometimes arrives too late, and the conversion is lost with no error message a marketing manager would ever read.
Retroactive consent. A contact who refused advertising personalisation after converting must not go out in later imports. That assumes the export re-reads the consent state on every run, not once at set-up.
How to connect a source properly: six steps
Map what leaves. Not the source, the connection: the named list of fields, each with its purpose and destination. If nobody can produce that list, the import is not ready.
Qualify the legal basis, record by record. Consent, contract performance, overriding interest: the answer can differ by segment, and the export filter has to reflect it.
Update the information you give. A privacy policy that names the recipients, mentions the disclosure to the United States and the safeguard relied on, and explains how to withdraw consent.
Choose the technical route. Point-and-click direct connection for a simple need and a supported source, Data Manager API when you need frequency, volume or encryption with your own keys. In the second case, decide immediately where the keys live.
Set the deduplication rule. Transaction identifier, import window, authoritative source per conversion type. Written down, not agreed verbally.
Put it under watch. The data source diagnostic alerts added in February 2026 give a view of connection health. Monitor match rate as well, the 540-day and 100-member rule, and review access every time a provider changes.
How A-Track can help
We work on all three layers of this subject, because they hold together. Our tracking and compliance audit establishes what currently leaves your site and your tools, including imports set up by former providers and never documented. Our tracking and compliance offer covers designing the full chain, from the server-side container to the export filter on the CRM side, with the documentation that goes with it. Our annual maintenance takes on the watch for this kind of switch, of the sort that happened on 15 June 2026, and the regular checking of connections.
We work from Geneva, with Swiss SMEs and with web and marketing agencies for whom we take on the measurement and compliance side of their own clients' work.
Conclusion: where to start
Data Manager has made sending customer data to Google Ads simpler than it has ever been, and that is precisely why it deserves a guardrail. The tool will not ask whether you are allowed to send what you are sending.
The first step is short and you can take it this week. Open Data Manager in your Google Ads accounts, list the existing connections, and for each one ask three questions: which fields leave, on what legal basis, and who decided. If a single connection has no answer, you have your starting point. And if your last offline conversion import predates June 2026, check first that it still works.