Helpful Fivetran Tutorials
September 17, 2026 · SyncSpur Team
Marketing connectors behave unlike database connectors: the same rows keep changing as conversions land, so every sync re-reads a rolling window of history. Here is how Fivetran's ad platform connectors actually work, how to configure conversion windows and report tables without inflating MAR, and when to use GA4's native BigQuery export instead.
Read moreSeptember 16, 2026 · SyncSpur Team
Pointing an LLM at your raw warehouse is a demo, not a product. How to build an agent-facing layer on Fivetran-landed data: what MCP gives you, exposing pipeline freshness first, curating a described semantic model, and the scoped roles, row-level security, query caps and prompt logging that keep it safe.
Read moreSeptember 15, 2026 · SyncSpur Team
Salesforce is the connector that behaves least like a database. A practical guide to running it well with Fivetran: a dedicated integration user, sync-method choice, what really burns your 24-hour API limit, why formula fields go stale, how purged and archived records rot your row counts, and an object list that keeps MAR under control.
Read moreSeptember 14, 2026 · SyncSpur Team
Once more than one squad is creating connectors, the Fivetran account itself becomes the risk. A practical guide to structuring destinations and connection groups, mapping Fivetran roles to real job functions, wiring SSO and SCIM provisioning from Okta or Entra ID, and proving who changed what with audit logs and the Platform Connector.
Read moreSeptember 14, 2026 · SyncSpur Team
Fivetran does not remove rows from your warehouse when they disappear at the source - it flags them. That default is right for analytics and wrong for a right-to-erasure request. A practical guide to how deletes propagate through CDC and API connectors, what _fivetran_deleted really means downstream, and how to build an erasure workflow that survives an audit without triggering a full re-sync.
Read more