GA4 (not set): Causes and Fixes by Dimension

June 24, 2026

Author: Shusaku Yosa

GA4 uses (not set) when a selected dimension has no value. It is a symptom, not a diagnosis. Adding UTMs or waiting for processing is not a universal fix. Start with the exact dimension, identify the information it requires, and test whether that information reaches the correct property. This workflow separates acquisition data, landing pages, and custom definitions so that each investigation has a clear success criterion.

Distinguish (not set), (none), (other), and Unassigned

(not set), (none), (other), and Unassigned describe different situations. Treating them as interchangeable makes it easy to change the wrong setting. A missing page referrer, for example, does not automatically make source / medium (not set): campaign parameters can still identify the visit.

Label	Meaning and next check
(not set)	No value for the selected dimension; inspect collection, definition, and applicability
(direct) / (none)	No clear acquisition information; not limited to typed URLs
(other)	Values aggregated because an underlying table exceeded its row limit
Unassigned	Traffic does not meet a channel-group rule; inspect source and medium

Google’s dimension-specific guidance and its explanation of the (other) row describe separate mechanisms. A high-cardinality reporting problem does not necessarily mean that an event parameter failed to arrive.

Fix the report, scope, date range, and denominator first

Record the property, exact report and dimension name, date range, hostname condition, metric, and filters before troubleshooting. “Source” alone is not enough: First user source, Session source, and the unprefixed Source dimension answer different questions. Keep very recent observations separate from dates that have completed more processing.

Choose a denominator appropriate to the question. Landing-page missingness belongs to sessions. Missing article metadata should normally be assessed on the events expected to carry it. A small overall percentage can still matter if it affects a critical success event or a valuable customer journey. Prioritize business impact as well as volume.

Session source / medium: investigate session_start

For Session source / medium, the official missing-value documentation identifies missing session_start events as a diagnostic starting point. Reproduce a session-starting visit and inspect the destination property, session information, and the order in which collection begins. Document whether the implementation includes browser tags, server-sent events, or both.

Missing Google Ads-specific dimensions require a separate investigation of linking and click information. An untagged URL is not automatically a (not set) session. Also check the dimension’s scope: unprefixed Source / Medium concerns key-event attribution, and values can be absent for non-key events. The traffic-source scope documentation helps distinguish an inapplicable field from a collection failure.

Landing page: confirm that page_view exists

A landing page is based on the first page view in a session. A session without a page_view can therefore have a (not set) landing page. Do not stop at the hypothesis that the URL was unavailable when the session began; confirm whether the event itself exists and carries the expected page_location.

Test the actual entry page, including direct entry to a deep URL.
Separate initial loading from SPA route changes and browser collection from server sending.
Check the intended production hostname and destination property.
Before adding page views, check whether automatic collection already sends them.

Use DebugView as evidence about a controlled reproduction. It is not a substitute for the later aggregated landing-page report, and sending an extra page_view can introduce duplicate measurement.

Custom definitions: match collection, registration, and events

Custom reporting needs both collected data and the appropriate registered definition. In an illustrative design, article_view sends article_category with the value analytics. Register that exact event parameter with Event scope if it is intended as an event-scoped reporting dimension. Choosing User scope does not turn an event parameter into a user property.

Dimension	First evidence to inspect	Validation evidence
Session source / medium	session_start and session information	A controlled new session and its processed report row
Landing page	page_view and page_location	Entry-page event and later landing-page row
article_category	article_view payload and matching definition	DebugView value and event-filtered report
Unprefixed Source	Key-event applicability and attribution scope	A report restricted to applicable events

Other events may legitimately omit article_category. Filter to article_view before calling those omissions defects. Custom-definition reporting normally needs 24–48 hours after the relevant setup and collection; processing time does not repair an unsent value.

When consent implementation, redirects, or tag order is involved, preserve the consent state and transition sequence in the test record. Compare the states relevant to the implementation without disabling consent controls to make a missing-value percentage look better. Redirects that lose campaign information should be investigated separately from absent page views.

In a GTM implementation, distinguish the current Google tag from GA4 event tags. Follow Google’s DebugView instructions to select the test device and inspect event names, times, and values. If an event is absent there, check debugging and consent conditions before concluding that all production traffic is missing it.

Calculate missing-value rates and validate a change

Consider an illustrative dataset, not a benchmark. Forty missing landing pages out of 1,000 sessions gives 40 ÷ 1,000 × 100 = 4%. If 2,000 of 10,000 total events lack article_category, 20% is not automatically an implementation error rate. If 30 of the 600 article_view events expected to carry it are missing the value, the relevant rate is 30 ÷ 600 × 100 = 5%.

Save the original conditions, counts, and denominator.
Have the responsible person review the suspected cause and proposed change.
Reproduce the affected journey and check for duplicate sending.
Compare processed data over equivalent periods with the same conditions.
Record remaining gaps and journeys that could not be tested.

A configuration change does not promise to backfill historical missing values. Choose one dimension first, document the missing evidence, and define what would demonstrate a successful correction.