Data Residency
ABetterChoice's default — and currently only — deployment model is Warehouse Native. The practical meaning for data residency is: your raw event data and user-property data never leave the warehouse account you own.
What stays inside your warehouse
The following data is read in place by the platform's statistical engine via SQL and is never copied to ABetterChoice infrastructure:
- Raw exposure rows in your Assignment Tables.
- Raw business events in your Fact Tables.
- User-property rows in your User Property Tables.
- All derived intermediate result sets used to compute metrics.
Computation happens inside your warehouse. ABetterChoice pushes SQL down; the warehouse engine runs it; results are materialized into result tables in the same warehouse.
What crosses the boundary
Only the metadata required to run the platform crosses into ABetterChoice's control plane:
| Crosses the boundary | Stored where |
|---|---|
| Project, experiment, remote-config, audience, metric definitions | ABetterChoice control plane |
| Table mappings (schema names, column names, partition columns) | ABetterChoice control plane |
| Result-table pointers (database / schema / table identifiers) | ABetterChoice control plane |
| Aggregated metric numbers displayed on the Results view | ABetterChoice control plane |
| Member account information (email, display name, role) | ABetterChoice control plane |
Aggregated result numbers are summaries — the underlying per-user rows that produced them remain in your warehouse.
Region and supported warehouses
The single warehouse a project can connect to today is BigQuery. Databricks, Redshift and Snowflake are in active development. The region in which your project data lives is therefore the region you chose when you created the BigQuery dataset.
Switching a project to a different warehouse detaches every existing data source. Treat region selection as a one-time decision. See Connecting Your Warehouse for the connection limits.
Revocation
Because ABetterChoice reads through an IAM role that you grant, revocation is a single warehouse operation:
- In BigQuery IAM, remove the ABetterChoice service-account principal from the role you granted in BigQuery Connection.
- The next SQL push from the statistical engine fails authentication; experiments stop refreshing immediately.
- To complete the off-boarding, also delete the warehouse connection from Data Management → Warehouse Native inside the console.
What is not yet documented
The following items appear in some Warehouse Native discussions and are still under product clarification:
- A formal VPC-only / PrivateLink deployment posture between the control plane and your warehouse.
- A signed Data Processing Agreement (DPA) appendix that enumerates sub-processors.
- A self-service export of all metadata for a leaving customer.
If your security review requires any of these, reach out to support — they are tracked but not generally available.