Combining Robin Signals and GA4 in BigQuery for Hybrid Tracking

Traditional web analytics relies on cookie-based client-side tags, which are increasingly degraded by consent rejection rates of 30-40% and browser-level caps on cookie lifetime. A privacy-first environment calls for a hybrid tracking architecture. By combining the campaign attribution and ad-platform integrations of Google Analytics 4 (GA4) with the complete, cookieless telemetry of Robin Signals, organisations build a more complete analytical foundation. Unifying these parallel event streams in a central data warehouse such as BigQuery keeps total volume metrics accurate while preserving downstream conversion attribution.

Implementing this hybrid setup begins at the collection layer. On the client side, both tags run in parallel. Robin Signals captures consent-independent, cookieless telemetry (such as page views, scroll depth, and basic interactions) unconditionally, using cryptographic session hashes instead of persistent cookies. When a user grants consent, GA4 is initialised and a custom identity_link event fires. This event carries both the GA4 client_id and the Robin Signals hash as event parameters, linking the two sessions in BigQuery.

In BigQuery, a lookup table maps the GA4 user_pseudo_id (the client_id as it appears in the export) to the Robin Signals hash. This identity mapping is derived exclusively from consenting traffic. Because the Robin Signals hash is computed deterministically from first-party headers, it remains stable. The mapping query filters the identity_link events over a rolling window (e.g., 30 days) to keep the mapping fresh and performant, producing a compact key-value lookup.

To construct the unified event stream, use Robin Signals as the primary dataset, because it captures interactions regardless of consent. Left-joining this stream with the identity mapping table classifies each event as consenting or non-consenting. For consenting users, GA4 events are joined to enrich the journey with GA4's native attribution, ad-click parameters (such as GCLID), and audience signals. For non-consenting users, Robin Signals data serves as the source of truth for volume, correcting the consent bias and showing the true paths.

Unifying these event streams unlocks several advanced use cases. First, it allows the calculation of a true conversion rate across the entire traffic base, rather than an inflated rate computed only on consenting users. Second, attribution models (such as Markov chains or logistic regression) can use the unified stream to credit channels for conversions that occurred in the non-consenting cohort. Finally, the hybrid data provides a sounder foundation for training propensity models in BigQuery ML, preventing the feature distortion caused by missing data.