Why can GA4 hide rows in a low-volume referral report?

Definition: GA4 data thresholding is a system-defined privacy rule that can withhold report data when a row has few users or when the report contains sensitive dimensions or signals. A missing row in a referral report can therefore result from the data requested by the report and its volume, rather than from the referral source being absent.

Key facts

• Google's GA4 data thresholds documentation says thresholds protect against inferring individual identity or sensitive information from demographics, interests, or other signals.
• Google says a row with a small number of users may be excluded, especially when a report uses high-cardinality or custom dimensions. The Data API response exposes a subjectToThresholding field for this condition, as described in Google's reporting data expectations.
• The Data API quota documentation names userAgeBracket, userGender, brandingInterest, audienceId, and audienceName as potentially thresholded dimensions.

What does low volume have to do with a missing row?

Google's thresholding rule is designed to prevent a report viewer from inferring the identity or sensitive information of individual users. The help page says data may be withheld when a report, exploration, or API call includes demographic data or audiences defined using demographic data. It also says that a row containing search query information may be withheld when there are not enough total users.

The reporting data expectations page describes the row-level effect directly: a row with a small number of users may be excluded from the results. It says this is more common with high-cardinality dimensions or custom dimensions. A referral report can therefore show other source rows while a low-volume combination is absent when the requested report includes a threshold-sensitive field.

Which dimensions can be affected?

In the Data API, a dimension is an attribute used to describe or group data. Google's API dimensions and metrics reference groups dimensions around user and event information, audiences, campaigns, and advertising platforms, and distinguishes them from metrics such as users, sessions, engagement, conversions, revenue, and ecommerce measurements.

Google's quota documentation explicitly identifies five potentially thresholded dimensions: userAgeBracket, userGender, brandingInterest, audienceId, and audienceName. The same page says the thresholds prevent someone viewing a report from inferring the demographics or interests of individual users. The reporting expectations page also identifies high-cardinality and custom dimensions as contexts where a small-user row is more commonly excluded.

The documented examples concern demographic, audience, search-query, high-cardinality, and custom-dimension data. Google's pages do not publish a numeric minimum aggregation cutoff for a low-volume source row.

How can the data quality indicator confirm thresholding?

In the GA4 interface, the data quality indicator at the top of a report carries a thresholding message when one or more cards are limited. Google says the message reports that data will be displayed only when the cards meet the minimum aggregation thresholds. This is the interface signal to check beside an apparently incomplete table.

For a Data API report, Google's reporting expectations documentation says to inspect the subjectToThresholding field in the ResponseMetaData object:

subjectToThresholding

The same documentation describes a separate high-cardinality behavior: less-common values can be grouped into an (other) row, more commonly when a dimension has more than 500 unique values per day. That aggregation behavior is distinct from a row withheld by thresholding.

What can change the result?

Thresholds are system-defined and cannot be adjusted. Google recommends changing the date range to decrease the likelihood of thresholds appearing. Its help page says a narrow date range with low user or event counts may withhold data, while expanding the range may increase the number of users who triggered an event and make previously thresholded data visible.

Google also documents BigQuery as a separate export path, with an important qualification: Analytics does not export data from Google signals to BigQuery, and event counts per user may differ between Analytics and BigQuery.

Related terms

• GA4 traffic acquisition
• AI referral traffic
• How do I query AI referral sessions in the GA4 BigQuery export?
• Why is older AI referral data missing from GA4 Explorations?

FAQ

Does Google publish a threshold number for a missing source row?

The documentation describes minimum aggregation thresholds and low user or event counts. It does not publish a numeric cutoff for a low-volume source row.

Can you turn off GA4 data thresholding?

No. Google says data thresholds are system-defined and cannot be adjusted. You can change the date range, which Google documents as a way to decrease the likelihood of thresholding.

Is a missing row the same as an (other) row?

No. A thresholded row may be excluded when it contains a small number of users. A high-cardinality dimension may instead group less-common values into (other).

Sources

1. Google Analytics Help, "[GA4] About data thresholds", undated
2. Google for Developers, "Reporting data expectations", 13 October 2026
3. Google for Developers, "Data API limits and quotas", 13 October 2026
4. Google for Developers, "API dimensions and metrics", 13 October 2026

---

Check how a page reads to answer engines. Open the GEO score analyzer