Set up your first reverse ETL sync without breaking prod

Set up your first reverse ETL sync without breaking prod
Warehouse-to-SaaS pipelines, done in the right order.

1 — Pick one model and one destination for v1 (e.g. churn-risk score to your CRM).
Why: scoping to a single field is the difference between a launch and a six-month project.

2 — Materialize the model as a table, not a view.
Why: live views re-query on every sync and will throttle your warehouse credits.

3 — Set the sync to 'upsert', never 'mirror', for your first run.
Why: mirror mode can null out fields the destination owns; upsert only touches yours.

4 — Define a primary key the destination already recognizes.
Why: without a stable match key, every sync creates duplicate records.

5 — Start at hourly cadence, not real-time.
Why: most activation use-cases don't need streaming, and hourly halves your row costs.

That's the stack for this week. Forward to a teammate.