Skip to content

Warehouse Integrations

Warehouse connectors are data sources. Adwize reads the streams you enable, copies them into Adwize storage, then runs monitoring and insights. This is not a vendor export API (Google Ads reporting, GA4 Data API) and not a full-project dump.

You pick a connector, grant access, detect objects, enable streams, and set a pull schedule. The first successful ingest uses the first-load window. Later runs start from that stream’s last day minus a 3-day overlap for late data.

Adwize can also write destinations on BigQuery and Cloud Storage if you turn that on. Snowflake, Postgres, MySQL, DuckDB, Azure Synapse, and Microsoft Fabric are source-only.

Shared steps

  1. In Adwize, go to Settings > Integrations > Add source and pick a warehouse.
  2. Grant access (IAM, JSON key, dedicated user, or MotherDuck token — depends on the connector).
  3. Check connection, then Detect streams.
  4. Enable the objects you want pulled. Save mapping.
  5. Set the account schedule (manual / daily / weekly) at the top of Integrations, and the first-load window on the connection. Manual does not pull until you change it.

Permissions

We read mapped tables or prefixes only. We copy enabled streams into Adwize. We do not dump the whole project, and we do not call Ads or GA reporting APIs to fill a warehouse.

BigQuery

Grant Data Viewer and Job User on each selected dataset (IAM on Adwize’s service account, or paste a JSON key you control). Map datasets and tables. Event history is the mapped BigQuery export. The Google Analytics connection is Admin config. Mapping events_YYYYMMDD does not link that table to the connected property. Ads spend belongs here as an export you already own, not via the Ads tool connector.

If you enable destination, Adwize can write tables back into BigQuery.

Cloud Storage

Grant Object Viewer on each selected bucket. Map bucket and prefix. Adwize copies enabled prefixes; it does not list every bucket in the org.

If you enable destination, Adwize can write objects back (CSV under a prefix).

Snowflake

Create a dedicated read-only user. Grant USAGE on the warehouse, database, and schema, and SELECT on the tables Adwize should sync. Do not use a personal login. Adwize does not write back to Snowflake.

PostgreSQL

Create a dedicated read-only user with SELECT on the tables Adwize should sync. Do not use a personal login. Adwize does not write back to Postgres.

MySQL

Create a dedicated read-only user with SELECT on the tables Adwize should sync. Do not use a personal login. Adwize does not write back to MySQL.

DuckDB

Point Adwize at a DuckDB file the worker can open, or a MotherDuck database with a token. Adwize copies enabled streams from that file or database only — not other files on the host.

Azure Synapse Analytics

Create a dedicated read-only SQL user on the dedicated SQL pool or serverless SQL endpoint. Grant SELECT on the tables Adwize should sync. Do not use a personal login.

In Integrations, enter the SQL endpoint, database, schema, and user, then the password. Check connection, detect tables, and enable the streams to pull. The account schedule copies those streams. Adwize does not write back to Synapse.

Allow the Adwize worker to reach the endpoint (firewall or private connectivity). The worker uses ODBC Driver 18 and encrypts the connection.

Microsoft Fabric

Fabric Warehouse SQL endpoints accept Microsoft Entra authentication, not a SQL password. Create an Entra application and grant it read access to the warehouse. Paste the SQL endpoint, warehouse name, schema, tenant ID, client ID, and client secret.

Check connection, detect tables, and enable streams. The account schedule copies those streams. Adwize does not write back to Fabric, and this connector does not read OneLake files.