Here is the correct logic for calculating the Google Analytics 4 ‘Landing Page’ dimension in BigQuery, which also takes (not set) landing pages into account:
A landing page is identified by the ‘entrances’ event parameter, which is set to 1 for the first ‘page_view’ event of a session.
For subsequent visits to the same page in the same session, the ‘entrances’ parameter does not appear in the ‘event_params’ array.
The landing page is tied to a unique session and a session is identified by a unique combination of ‘user_pseudo_id’ and ‘session_id’.
Therefore, the landing page is unique for each unique combination of ‘user_pseudo_id’ and ‘session_id’.
Even if a user visits the same page multiple times in the same session, it will count as the landing page for the session only once.
Note: You may sometimes see rows where entrances = 1 for ‘page_location’ where both ‘user_pseudo_id’ and ‘session_id’ are null.
This could be due to bot traffic, consent mode or some data collection issue. You can ignore such rows if there are not many of them.
If the same user starts a new session (with a new ‘ga_session_id’), the landing page is calculated separately for that session.