GA4 user-provided data: what the new diagnostic is telling you
On 11 November 2026 Google added a diagnostic for user-provided data in Google Analytics. It watches whether your property receives the hashed identifiers you promised to send, and whether the quality of what arrives is good enough to be matched against Google accounts.
That second half is the useful part. A broken user-provided data setup has always failed quietly. The tag fires, the request returns 200, nothing appears in any error log, and the only visible symptom is a match rate that drifts down over months while enhanced conversions and Customer Match underperform. Now the property says so directly, in two specific warnings.
What the diagnostic checks, and where it appears
Alerts show up in the property's data collection settings, under Admin, then Data collection and modification, then Data collection. The same items also surface in the Data streams panel and in Task Assistant, which is where most people will meet them first.
There are two headline warnings, and they mean different things.
- Your property is not yet receiving UPD from a Google tag or Data Manager API. Collection is switched on and the terms are acknowledged, but no identifier has arrived.
- Your property's UPD has a low PII rate. Identifiers are arriving, but the share of events carrying usable personal data is too low to help matching.
The first is a wiring fault. The second is a quality problem, and it has six documented causes that look identical from the outside.
What user-provided data actually is
User-provided data is a set of hashable fields, meaning email address, phone number and postal address, that you send with your events so Google can match them against signed-in accounts. It powers three things: enhanced conversions for your Analytics conversions, Customer Match support for audiences exported to linked advertising products, and demographic and interest reporting built on first-party data.
It is not the same thing as User-ID. A User-ID is an identifier you assign to a signed-in user, typically from a CRM, and it stitches sessions and devices together inside your own property. User-provided data leaves your property and gets matched against Google's data. Google recommends sending both, because a User-ID survives a device change while an email address survives a CRM migration.
Three constraints matter before anyone turns this on. User-provided data collection requires a link between the Analytics property and a Google Ads account, it is not available for properties in the Health industry category, and the feature itself is still labelled an open beta, so the behaviour is subject to change.
The first warning: nothing is arriving
Collection is enabled in the property settings, the policy is accepted, and no data reaches Analytics. In practice this comes from one of three setup gaps.
The likely one is that the tag was never updated. Switching the feature on in the Analytics interface does not make the tag send anything. The tag has to carry a user_data parameter, either through a User-Provided Data variable in Tag Manager or through a manual call in gtag.js. Enabling the property setting is half the job, and it is the half that produces no traffic.
Second, the automatic detection checkbox. Analytics offers "Collect automatically detected user-provided data", which asks Google to find identifiers on the page itself. That option only works if automatic collection is also enabled in the Google tag. Turning on one without the other is a common state, and it looks like a working configuration in both interfaces.
Third, a scope gap. A tag updated on the checkout page but not on the lead form, or updated in the web container but not in the server container that forwards events, sends identifiers from some journeys and none from others. If the only tagged step is a page most visitors never reach, the property has been set up correctly and still receives almost nothing.
Google's fix list names the three collection routes explicitly: automatically collected methods, CSS selectors, and manual code added to the page through the Google tag, Tag Manager or the Data Manager API.
The second warning: data arrives, but not enough of it
A low PII rate means the identifiers that reach Analytics are too sparse to match well. Google groups the causes into six, and the order below is the order worth investigating them in.
Automatic collection is doing the work
Google states plainly that manual in-page code produces a higher PII rate than automatic collection. Automatic detection looks for form fields it recognises and gives up where the markup is unusual. Manual configuration through a User-Provided Data variable, or an explicit gtag call, tells the tag exactly where the values live. If the property relies on automatic detection alone, this is the first thing to change.
The parameter is missing on some events
Some events carry no user data fields at all, which means the parameter was not transmitted when that particular tag fired. The cause is usually a second tag or a second trigger added after the original setup, sometimes by a different agency, sometimes by a plugin. The check is per template, not per site: open each page where a conversion tag can fire and confirm the parameter is present on every one.
The fields are sent but empty
This is the most common of the six and the least obvious. The parameter is there, the fields are there, and the values are blank. Three reasons recur. The data genuinely is not on the page, because a signed-in customer never re-entered their details and the identifier was captured on an earlier step. The CSS selector points at an element that exists on desktop but not on mobile, or that renders after the tag fires. Or the selector matches a wrapper whose text content includes a label, so the value arrives as something like "Email address" instead of an address.
Google's advice is to test each event page in Tag Assistant and to repeat the test across browsers and devices, which is the only way to catch a selector that works on your laptop and fails on a phone.
The format fails validation
An email address without an @ symbol, or a phone number containing letters, fails formatting validation before hashing. This is not a matching problem that surfaces later. The value is rejected at the door.
The address is incomplete
One identifier from the three categories is required, so email, phone or address, and email is the preferred one. Address matching then needs its fields in a specific order: first name, surname, street address, city, region, country, postcode. Street address, city and region are optional keys that improve match rates, but the rest are expected. An address record that arrives missing a surname or a country code is not usable.
Individual fields are malformed
Separate from the general validation, Google lists field-level rules. Email should be an ordinary address format. Phone numbers follow the E164 standard with a leading plus sign and no non-numeric characters. First name and surname should be alphabetic strings. Country should be a two-letter ISO 3166-1 alpha-2 code, so GB rather than United Kingdom. Postcode should carry the number of digits the country expects.
Reading the error codes in Tag Assistant
The diagnostic page names the string to look for, and it is the most practical thing in the whole release. In Tag Assistant, the user data parameter appears in a form that starts tv.1~em. followed by the hashed value. If what follows that prefix is not a hash, it is one of five codes, and each one points at a different fault.
- Empty, with only
tv.1visible. The parameter exists and the email field is blank. Either the page has no data or the setup is wrong. - e0. The value failed formatting validation. An email without an @ symbol, a phone number containing letters.
- e1. Unsupported browser. Google expects this to be rare, and it suggests retrying the conversion in an updated Chrome.
- e2. SHA256 failed, usually because unexpected characters reached the hashing function.
- e3. The page is not served over HTTPS. User data is only hashed on secure pages.
That list turns a vague quality complaint into a specific fix. e0 is a formatting bug in the site. e2 is usually a value that arrived with whitespace or a hidden character attached. e3 means a test environment or a forgotten HTTP page is sending conversion data, which is worth knowing for reasons beyond measurement.
The consent and governance layer
Everything above is mechanical, and the mechanical part is the easy part. Sending hashed identifiers to Google is a data protection decision, and it belongs to whoever owns the privacy notice and the lawful basis, not to whoever owns the container.
Two details from Google's documentation shape that conversation. First, hashing on the web can be done in the browser by Google, or you can hash values yourself before sending them, and Measurement Protocol always requires you to hash. Second, retention of user-provided data follows the property's user data retention setting, which is two or 14 months, configured under Admin, then Data collection and modification, then Data retention.
There is also a switch with an unusual shape. Acknowledging the user-provided data policy is permanent, and it cannot be withdrawn. The collection checkbox beside it can be unchecked, and unchecking it stops the property processing user-provided data. Your tag may still collect it, but the property will not process it. There is a second control at the tag level, "Allow user provided data capabilities" in the Google tag settings, which stops the tag sending it in the first place.
Practically that means the off switch exists at two layers, and a client who asks to stop should have both checked. Leaving the tag sending identifiers that the property then discards is a data flow nobody wants to explain in a review.
One timing expectation to set with them as well: improvements to modelling and attribution from enhanced conversions can take up to a month to appear in Analytics and Ads reports. A fix that works will not show up tomorrow, and a client watching a match rate daily will conclude it failed.
A working order for the first pass
- Confirm the property is linked to a Google Ads account, because nothing else works without the link.
- Check the first warning before the second. A property receiving nothing has no quality problem to diagnose.
- Verify the tag carries the
user_dataparameter on every conversion template, not one. - If automatic detection is on, enable automatic collection in the Google tag as well, or switch to manual configuration.
- Open each conversion page in Tag Assistant and read the string after
tv.1~em.. Codes first, coverage second. - Test the same pages on a phone. Selector failures that appear only on mobile are invisible from a desk.
- Fix formatting at the point where the value is read, not in a spreadsheet afterwards.
- Write down the lawful basis, the retention setting and the two off switches before the setup goes live, not after a subject access request.
The diagnostic does not make user-provided data collection any easier to implement. It makes failure visible, which is the part that was missing. A property that receives nothing now says so in the Admin panel instead of waiting for someone to notice that enhanced conversions stopped matching.
Seeing a user-provided data warning in your property? Book a free tracking audit and we will trace the identifier from the form field to the tag, read the error codes with you, and confirm what Google actually receives.