Document all active Informatica workflows, mappings, and source/target connections. Identify which workflows are pure data movement (candidates for Fivetran) vs. those with complex transformations (keep in Informatica or refactor). Export connection metadata. Note any custom transformations, aggregations, or data quality rules embedded in mappings.
⚠️ Watch Out For:
- Informatica's metadata is complex—allow time to navigate the repository
- Custom transformations (REST API calls, SQL, JavaScript) may not map to Fivetran—identify these early
Sort workflows into two buckets: (1) candidates for Fivetran (simple source→warehouse syncs) and (2) keep in Informatica or refactor elsewhere. Pure data movement workflows with light transformation are ideal migration targets. Complex transformations belong in dbt, Informatica, or specialized tools.
⚠️ Watch Out For:
- Overestimating what Fivetran can handle leads to rework—be conservative about transformation complexity
- Some workflows may have hidden dependencies on Informatica's session logic—trace these carefully
For Fivetran-bound workflows, sketch how data will flow: source → Fivetran → warehouse → dbt (for transformation). Separate concerns: Fivetran handles loading, dbt handles transformation, Informatica (if needed) handles governance/lineage. Document the new architecture.
⚠️ Watch Out For:
- Mixing Informatica and Fivetran can be confusing—clearly document which tool handles what
- Data quality checks may need to move from Informatica mappings to dbt tests
Set up Fivetran account and workspace. Create connections for your data sources (Salesforce, Postgres, Snowflake, etc.). Create connections for your warehouse destination. Test each connection with a simple query.
⚠️ Watch Out For:
- Fivetran requires explicit IP whitelisting for some sources—verify network rules are in place
- Warehouse credentials must have appropriate schema/table creation permissions
Select the simplest Informatica workflow (ideally a single source→table mapping). Create the equivalent Fivetran connector. Configure table selection, column filtering, and sync mode (full vs. incremental). Test the sync. Compare outputs with the original Informatica run.
⚠️ Watch Out For:
- Column ordering and data types may differ between Informatica and Fivetran—normalize if needed
- Informatica's incremental logic may differ from Fivetran's cursor-based approach—test carefully
If your migrated workflows include transformations, move them to dbt. Create dbt models that reference Fivetran-loaded tables. Test dbt runs. Set up dbt to run after Fivetran syncs complete (via Fivetran webhooks or orchestration tool).
⚠️ Watch Out For:
- dbt models need to be written from scratch (no auto-conversion from Informatica mappings)
- Schedule dbt runs after Fivetran syncs; timing mismatches cause data freshness issues
Progressively migrate remaining Informatica workflows to Fivetran. Start with simple ones and work toward complex. For each, validate data matches the original Informatica output. Document any custom logic that couldn't be migrated.
⚠️ Watch Out For:
- Mid-size migrations can plateau—early workflows are easy, later ones may have hidden complexity
- Some Informatica-specific patterns (like dynamic file processing) don't map to Fivetran—plan alternatives
Once all data movement is in Fivetran, disable or remove the corresponding Informatica workflows. Keep Informatica running for governance/lineage or complex transformations if needed. Document which Informatica components are still in use.
⚠️ Watch Out For:
- Don't delete Informatica workflows immediately—archive them for 2-4 weeks for recovery if needed
- Informatica licenses are often bundled—verify you're freeing up capacity if you reduce deployment size
Before final cutover, run both Informatica and Fivetran pipelines in parallel for 1-2 full cycles. Compare record counts, data accuracy, and timing. Validate that downstream reporting and dashboards match expectations with Fivetran data.
⚠️ Watch Out For:
- Timing differences (Informatica runs at midnight; Fivetran at 6am) make comparison tricky—align schedules temporarily
- Historical data mismatches often stem from incremental sync cursor differences—investigate thoroughly
Finalize cutover by retiring old Informatica workflows from the critical path. Update SLAs and alerting for Fivetran. Monitor costs for first month (MAR-based pricing). Optimize sync schedules and column selection to manage costs. Document lessons learned.
⚠️ Watch Out For:
- First month of Fivetran costs can surprise if you don't optimize large tables—review data volume closely
- Team may still reference Informatica documentation—update wikis and runbooks proactively
