GA4 Source / Medium Explained: Scope, Referrers, and Direct Traffic

July 7, 2026

Author: Shusaku Yosa

GA4 source / medium describes the origin and method associated with traffic. A page referrer describes a preceding URL. These concepts are related, but they are not identical. First-user acquisition, session acquisition, and key-event attribution also answer different questions. Choose the scope before interpreting a number, then use the appropriate report and evidence to investigate unexpected values.

Separate source, medium, and page referrer

Source names the origin, such as google or newsletter. Medium describes the method, such as organic or email. Page referrer is based on page_referrer and describes the preceding page URL when that information is available. A processed acquisition label is not simply another name for the browser’s preceding URL.

Field	Illustrative value	Interpretation
Source	newsletter	A campaign’s named origin
Medium	email	Its specified traffic method
Session source / medium	newsletter / email	The acquisition source associated with a session
Page referrer	A preceding URL or no value	Context associated with the event

A tagged email link can identify source and medium without a browser-provided referrer. Conversely, an internal referrer URL does not necessarily establish a new acquisition source. Confirm exact names in the official dimensions reference.

Choose first-user, session, or key-event attribution scope

Use First user source / medium to ask how a user was first acquired. Use Session source / medium to ask which source acquired a session. Unprefixed Source / Medium concerns event-scoped key-event attribution, so do not read it as an interchangeable session dimension.

Question	Dimension	Typical interpretation
How was the user first acquired?	First user source / medium	User analysis by first acquisition
Which source acquired sessions?	Session source / medium	Session and engaged-session comparison
Which touchpoints received key-event credit?	Source / Medium	Credit under the selected reporting attribution model

Google’s scope documentation explains that changing the reporting attribution model does not change first-user or session-scoped dimensions. User, session, and event metrics are also different quantities; none should automatically be relabeled as registered people.

Find Session source / medium in Traffic acquisition

Open Traffic acquisition and select Session source / medium as the table dimension. Its location can vary with the report collection, so use the report and dimension names rather than relying on a single menu path. Confirm the date range and property timezone, then choose appropriate metrics such as Sessions and Engaged sessions.

Keep Session source / medium and Sessions fixed when evaluating acquisition.
Inspect the hostname condition if several hosts use the property.
Record an Organic Search condition when analyzing that channel separately from all traffic.
Use a separate event-oriented table with Page referrer and event_name to inspect preceding pages.

Event count is not a count of arrivals. One session can include several views and clicks, and the same user can return. Do not report events grouped by a session source as the number of people that source brought.

Read a multi-visit example without mixing scopes

Consider an illustrative journey where the same user remains identifiable. They first arrive through google / organic, return in a separate newsletter / email session, and later type the URL. Cookies, available signals, and lookback conditions affect attribution, so this is a reading exercise rather than a guaranteed report output.

Visit	First-user interpretation	Session and key-event interpretation
First organic visit	google / organic	The initial session is google / organic
Separate email visit	First acquisition remains google / organic	That session is newsletter / email
Later typed visit	First acquisition remains unchanged	Session attribution may use an eligible prior non-direct touchpoint; key-event credit is assessed under its own model

Session attribution uses non-direct last click with lookback conditions. A typed return therefore does not guarantee that Session source will be direct. Fractional key-event credit is a separate measurement from three visits by one identifiable user.

Distinguish direct, (not set), and Unassigned

(direct) / (none) is used when clear acquisition information is unavailable. Bookmarks and typed URLs are possible explanations, as are journeys that lose campaign information. An empty referrer can still coexist with a campaign source or applicable earlier non-direct attribution. See Google’s explanation of direct traffic.

(not set) means a value is unavailable for the selected dimension. Unassigned means channel-group rules have not classified the traffic. Inspect the source / medium strings, channel grouping, and page-referrer evidence separately. Lowering the direct percentage by itself is not a useful measurement objective.

Validate UTM and auto-tagged landing URLs

For email and external campaigns you control, define consistent utm_source, utm_medium, and utm_campaign values. Standardize spelling and capitalization. Follow the original link through shorteners and redirects to the final page, checking that the intended values survive. For auto-tagged Google Ads traffic, inspect click information and linking separately rather than layering manual UTMs without a plan.

Use Google’s campaign URL guidance and inspect Page location when validating the landing URL. UTMs can be omitted from Landing page + query string, so that field alone cannot prove that parameters were never sent. Avoid UTMs on internal navigation. In GA4, new campaign information during an existing session does not automatically start a new session or replace that session’s acquisition source.

Separate cross-domain measurement from unwanted referrals

A referral domain is a diagnostic clue, not an instruction to exclude it. A blog linking to your site may be a valid acquisition source. A second measured domain may need identity continuity. A payment-provider return may need unwanted-referral handling. These are different decisions.

Observed domain	Next check	Risk of blanket exclusion
External publisher	Actual referring link	Hides a useful acquisition source
Another measured domain	Shared stream, tags, and _gl	Ignoring the referrer does not connect identifiers
Payment or authentication service	Return path and referral-handling purpose	May conceal a separate success-measurement issue
Your own site	Tag coverage and cookie-domain settings	Skips the implementation diagnosis

Unwanted referrals changes referral treatment; cross-domain measurement connects eligible measurement across domains. Do not use one as a substitute for the other.

Document report conditions and measurement limits

Keep the date range, timezone, hostname, country and device conditions, scope, metrics, and tag or consent change dates with the analysis. A path-only report does not prove that all observations came from the intended production host. Document journeys where collection or identification is unavailable.

Before evaluating registrations or purchases, verify success conditions, duplicate sending, counting method, and authentication or cross-domain transitions. Unverified event totals are not registered people. Record unavailable outcome measurement as unmeasured, then choose the acquisition or attribution question that the available evidence can actually answer.