Where the numbers come from and what not to trust

Two places hold the data. GA4 is for looking; BigQuery is for answering questions GA4 cannot.

GA4. The corporate property sees every venue and every event. Every event carries the custom dimensions (location_id, location_name, partner_id, partner_name, page_type, venue_slug, and the rest), so any report can be sliced by venue or partner. Partner properties see only that partner's venues.

BigQuery. Every GA4 event, one row each, with all parameters. Use it for funnels by venue, anything per user rather than per session, joins between website and checkout events, and experiment reads. Dataset analytics_551852403 in project agi-mkt-analytics.

Things people get wrong:

• Revenue is value, not roller_amount_charged. The second is the deposit.

• Consent. Visitors who have not accepted analytics still send events, but without a user ID. In BigQuery they show as user_pseudo_id null and privacy_info.analytics_storage = 'No'. They count as events, not as users, and cannot be joined into a funnel.

• begin_checkout is not "slide-out opened" for every venue. See the events article.

• Event counts versus users. initiate_checkout can fire several times per visitor. For conversion rates, count distinct users at each step.

A short set of standard queries (funnel by venue, split by partner, exposures by variation) lives in the BigQuery article and should be the starting point for any ad hoc question.