session_start Is Not the First Meaningful Event in a Session

In the GA4 BigQuery export, session_start is often treated as the moment a visit begins. The name invites that reading, but relying on it alone distorts entry analysis and funnel timing.

session_start is a housekeeping event: it records that GA4 opened a new session after inactivity or on a new visit. It says nothing about what the user came to do. A user who follows an email link that triggers an add_to_cart API call before any page_view renders still generates a session_start, but the add_to_cart, if it is tracked (for example server-side with Robin Signals), is the event that carries the intent and value of that session.

Timing matters too. When session_start precedes the first page_view or view_item by several seconds, measures such as time on the first page or immediate conversion rate shift. The effect is largest in single-page applications and fast journeys, and it changes which touchpoint is credited as the entry.

A better approach is to model a session_first_event_timestamp in dbt. Take MIN(event_timestamp) across the events in each session, keyed on user_pseudo_id plus ga_session_id (the session ID alone is not unique across users) and excluding housekeeping events such as session_start and first_visit, then record the event name at that timestamp. Entry analysis can then group by the actual first interaction. In some implementations that first event is a click, a scroll or a custom event rather than the assumed page_view.

Where a session starts feeds everything downstream, from lifetime value models to causal estimates of conversion rates. Modelling the first meaningful event explicitly, instead of inheriting session_start, keeps that foundation aligned with what users actually did.