Mathilde Chevalier
04/28/2026, 10:31 AMkapa.ai
04/28/2026, 10:31 AMid in BigQuery.
Here are a few things to check based on the Airbyte documentation: How Full Refresh - Overwrite + Deduped works In this mode, each sync replaces all data in the destination table and deduplicates the result using a primary key. The deduplication is determined by sorting the data using the cursor field and keeping only the latest row per primary key. [Full Refresh Overwrite+Deduped] Possible reasons for duplicates 1. Primary key not set correctly: The deduplication requires a primary key to be defined. If the id field is not configured as the primary key in your stream's schema/mappings, deduplication won't work as expected. Make sure id is set as the primary key for the users stream. [Sync Modes]
2. Sync resumption: The documentation notes that Full Refresh syncs may be resumed (e.g., to recover from transient errors), and when this happens, duplicates can occur. This is actually one of the reasons Overwrite + Deduped is recommended over plain Overwrite ā to handle such cases. [Full Refresh Overwrite+Deduped] Recommended action ⢠Go to the Schema tab of your connection and verify that id is explicitly set as the primary key for the users stream.
⢠Also confirm that a cursor field (e.g., regtime, which appears in your query) is configured, as deduplication sorts by the cursor field to keep the latest record.
If the primary key and cursor field are correctly configured and you're still seeing duplicates, this may be a bug worth reporting to the Airbyte team with your version details (2.1.0). š 0 button š 0 button ā¹ļø Tip: Mention @kapa.ai in the thread for follow-ups or DM the app for 1-to-1 help.Mathilde Chevalier
04/28/2026, 10:32 AMMathilde Chevalier
04/28/2026, 10:35 AMkapa.ai
04/28/2026, 10:35 AMid field) to identify unique records
⢠Using the cursor field to determine which version of a record is the most recent (keeping only the latest row per primary key) [Full Refresh Overwrite+Deduped]
If no cursor field is configured, Airbyte may not be able to properly determine which duplicate row to keep, which could explain why both rows are being retained. What to try 1. Go to the Schema tab of your connection
2. Expand the users stream
3. Set a cursor field ā based on your earlier query, regtime looks like a good candidate (it appears to be a registration timestamp)
4. Save your changes ā you may need to trigger a Clear and re-sync to apply the changes cleanly [Select Sync Mode]
Note that cursors and primary keys defined by the source connector are automatically selected and cannot be changed, but if they are not source-defined, you can set them manually in the Schema tab. š 0 button š 0 button ā¹ļø Tip: Mention @kapa.ai in the thread for follow-ups or DM the app for 1-to-1 help.