Different processing methods for Google Analytics data in BigQuery can result in discrepancies between daily export and Fresh Daily. Discrepancies of less than 2% are expected between the two export types.
Differences in event counts
| Feature | Fresh Daily export | Daily export | Differences in export types |
|---|---|---|---|
|
Events per user limit (typically encountered when missetting User Id)
|
Enforce events per user limit continuously. Events exceeding the limit may be dropped throughout the day. | Enforce events per user limit after processing all events for the day. | Different events may be dropped due to the timing of enforcement. |
|
Sub and rollup property events filter
|
May over-filter or under-filter due to limited session context. | Has full session context for filtering. | Potential for over-filtering or under-filtering in streaming processing. Sub and rollups are not covered by the Fresh Daily SLA . |
|
Spam events filter
|
As spam filters are deployed throughout the day, some events will be exported before the filter is deployed. | Applies spam filters after all events are collected for the day. | More spam events might be present in streaming data. |
Differences in event dimensions and metrics
| Feature | Fresh Daily export | Daily export | Differences in export types |
|---|---|---|---|
|
Active user ID
|
Determined based on partial events data due to continuous processing. | Determined based on all events for the day. | Possible discrepancies in active_user_id dimension count. |
|
Audience evaluation
|
Evaluated without offline events and audience triggers. | Evaluated with all events, including offline events and audience triggers. | Differences in audience evaluation results. |
|
Attribution results
|
Search Ads 360 (SA360) and app attribution results may not be present | Contains SA360 and app attribution results. | Some widening sources are not available. Discrepancies in attribution are expected given widening differences. |
|
Server-side state
|
Partial data due to continuous processing every 30-60 minutes. | All events from the current data. | Differences in dimensions and metrics that rely on server-side state like engaged_sessions and last_purchase_date. |
