Start by documenting every active connector in your Stitch workspace. Export a list of all source connections, connector types, replication frequencies, and destination warehouse connections. Note any custom transformations or dbt runs triggered by Stitch. This inventory becomes your migration checklist.
⚠️ Watch Out For:
- Don't forget archived or disabled connectors—you may want to migrate those too
- Stitch's webhooks and post-load dbt runs aren't always obvious in the UI
For each Stitch connector, identify the Fivetran equivalent. Most major sources (Salesforce, Stripe, HubSpot, Postgres, MySQL) have direct equivalents. For any Stitch connectors without a Fivetran match, research alternatives: some sources might be available through Fivetran's API connector or community connectors.
⚠️ Watch Out For:
- Some Stitch connectors (particularly older, less popular ones) don't have Fivetran equivalents
- Fivetran's column selection and sync options may differ from Stitch—table-level selection vs. column-level
Create a Fivetran account and configure your workspace. Set up team members with appropriate roles (Admin, Creator, Viewer). Configure SSO if required. Enable audit logging for compliance. Create destination connections for each target warehouse you use with Stitch.
⚠️ Watch Out For:
- Fivetran's role-based access control works differently than Stitch—plan team permissions carefully
- Warehouse connection setup requires network/firewall adjustments for some warehouse types
Create and configure your first source connector in Fivetran. Start with a non-critical source to test the process. Authenticate the connector with source credentials. Configure table selection and column settings. Set initial sync parameters (full refresh vs. incremental).
⚠️ Watch Out For:
- Fivetran requires explicit table selection—syncing 'all tables' defaults are more conservative than Stitch
- Some sources require IP whitelisting; ensure Fivetran IPs are added to your source firewall rules
Trigger test syncs on your first few connectors. Compare record counts and sample rows between Stitch and Fivetran. Verify that incremental syncs are working correctly. Check for any schema mapping issues or data type differences. Run SQL queries on both destinations to validate matching counts.
⚠️ Watch Out For:
- Fivetran's incremental sync logic may differ from Stitch—carefully review high-water mark columns
- Data type conversions (especially timestamps and JSON) can introduce subtle mismatches
Map Stitch's replication frequency settings to Fivetran's scheduler. Fivetran uses 'sync frequency' (hourly, daily, etc.) rather than Stitch's minute-based intervals. Set appropriate sync windows if you have off-peak sync requirements. For real-time sources, configure CDC settings if applicable.
⚠️ Watch Out For:
- Stitch allows 5-minute sync intervals; Fivetran's minimum is hourly—this may increase latency
- Fivetran's 'Connectors Manager' for bulk schedule changes is not as granular as Stitch's per-connector UI
If you used Stitch's dbt integration, reconfigure your dbt runs to trigger from Fivetran's webhooks instead. Alternatively, move to Fivetran's built-in transformation (DBT) or external orchestration tools (Airflow, Dagster). Update any downstream dependencies that assumed Stitch's loading schedule.
⚠️ Watch Out For:
- Timing assumptions matter—moving from 5-minute to hourly syncs can break downstream jobs
- Fivetran's transformation output table naming differs from Stitch—update dbt refs accordingly
Once all connectors are running in Fivetran and passing validation, pause or disable the corresponding Stitch connectors (don't delete immediately). Run parallel syncs for 1-2 full cycles to ensure Fivetran data matches Stitch. Update any analytics tools, dashboards, or reports to point to Fivetran-loaded tables if schema names differ.
⚠️ Watch Out For:
- Table naming in Fivetran defaults to source_schema.table; adjust if you prefer a different convention
- If you switch off Stitch before full validation, you risk data gaps for tables with long sync windows
After 1-2 weeks of Fivetran operation, review your consumption costs (MAR—Monthly Active Rows). Identify heavy-syncing tables and consider adjusting sync frequency or column selection. Configure Fivetran alerts for sync failures. Compare actual costs against your budget and Stitch's previous spend.
⚠️ Watch Out For:
- Fivetran charges by MAR (Monthly Active Rows); syncing large tables daily can be expensive
- Some sources (like Salesforce) have high per-connector costs regardless of data volume—plan accordingly
After 2-4 weeks of successful Fivetran operation, delete Stitch connectors and disconnect the destination warehouse. Archive or delete any Stitch-specific documentation. Update team runbooks and data lineage documentation to reference Fivetran instead. Notify analytics and data engineering teams of any table name or refresh schedule changes.
⚠️ Watch Out For:
- Don't decommission Stitch too quickly—historical backups or audit trails may require access
- Team members may still reference Stitch documentation; ensure you update internal wikis and training
