Reverse ETL is the pattern of moving data out of your data warehouse and into operational tools like CRM, ad platforms, and sales engagement. It is the inverse of traditional ETL, which moves data into the warehouse for analysis. Reverse ETL is what makes a warehouse usable as the system of record for go-to-market motion.
Why it exists
For most of analytics history, the warehouse was a dead-end. Data engineers loaded customer behavior, product usage, and revenue into Snowflake or BigQuery, analysts built dashboards, and the operational tools (Salesforce, HubSpot, Marketo) lived in a separate world. When marketing wanted “users who logged in three times last week,” someone had to write a custom integration or wait for a CSV.
Reverse ETL closes that loop. Tools like Hightouch, Census, and Polytomic let you write SQL once against the warehouse, define a sync to a destination, and have the data appear as fields in Salesforce or audiences in Facebook within minutes.
When it matters
Reverse ETL is the right pattern when three conditions hold: the data you want to act on already lives in the warehouse and is hard to recompute elsewhere; the destination tool has a usable API or native connector; and the team operating the syncs is data-literate enough to own SQL models.
The classic use cases are product-qualified leads (sync product usage signals to CRM as lead scores), audience targeting (push warehouse segments to ad platforms), and customer health (sync support, billing, and product data to CSM tooling).
What it replaces
Reverse ETL competes with three older patterns: hand-written API integrations (slow, brittle), iPaaS like Zapier and Workato (good for simple workflows, expensive at scale), and packaged CDPs (more opinionated, less flexible). The composable CDP pattern is essentially reverse ETL plus an identity model.
Common pitfalls
- Syncing the wrong grain. Pushing every product event to Salesforce will destroy it. Aggregate to the right grain (account-week, user-day) before syncing.
- Letting models drift. A reverse-ETL sync depends on an upstream dbt model. When the model changes, the destination breaks silently. Add tests and ownership.
- Activating before consent. Reverse ETL respects whatever consent the warehouse knows about. If consent is not modeled in the warehouse, you are pushing non-compliant data.
Related
- Data warehouse vs CDP — the architecture choice
- CDP — the alternative pattern
- RevOps tech stack — where reverse ETL fits