Reverse ETL is the category of tools that does the exact opposite of normal ETL. Where Fivetran and Airbyte pull data from SaaS apps into the warehouse, reverse ETL pushes data from the warehouse back into SaaS apps. The same Salesforce account record that started in Salesforce, got synced to Snowflake, got enriched with product usage data, and got modeled by dbt — reverse ETL is what shoves the enriched version back into Salesforce so the sales rep can actually see it.

The simple way to think about it: the warehouse is where you compute the truth, and reverse ETL is the delivery truck that takes the truth to wherever it needs to be used. A modeled customer health score in Snowflake is useless if it stays in Snowflake. The customer success rep lives in Gainsight or HubSpot. The marketer lives in Iterable. The ad buyer lives in Google Ads. Reverse ETL is the wire that connects warehouse intelligence to the places work happens.

Where the Category Came From

Until about 2019, "operational analytics" — the idea of acting on warehouse data inside operational tools — was a mess of one-off Python scripts, hand-rolled API integrations, and Zapier hacks. Every data team had a folder of cron jobs that pushed CSVs to Salesforce or fired webhooks at HubSpot. These scripts broke constantly, had no observability, and were owned by one engineer who would inevitably leave.

In 2018-2019, two companies independently noticed the same opportunity and built the same product: a managed pipeline that takes a SQL query against your warehouse and syncs the result to a SaaS destination, with idempotency, error handling, field mapping, and a UI that didn't require a Python engineer.

• Census, founded in 2018 by Boris Jabes and Bastin Gerald, came out of stealth in 2019.
• Hightouch, founded in 2018-2019 by Tejas Manohar, Kashish Gupta, and Josh Curl, also launched in 2019.

The term "reverse ETL" was coined around the same time — often attributed to Astasia Myers at Redpoint Ventures, who wrote some of the first analyst pieces on the category, and to the founders of both companies who were trying to explain what they did. The name is technically a misnomer (it's really "ELT in reverse"), but it stuck because it told you exactly what the product did in two words.

By 2021, reverse ETL was the hottest new category in the data stack. By 2024, it had become so obviously useful that the warehouses themselves started absorbing the functionality. Snowflake announced reverse-ETL-style features. Databricks built outbound connectors. Salesforce shipped Data Cloud, which is essentially a reverse ETL pipeline pretending not to be one.

What Makes a Reverse ETL Tool a Reverse ETL Tool

Three properties:

1. The warehouse is the source of truth. A reverse ETL tool reads from your data warehouse — not from raw production databases, not from a separate CDP, not from event streams. The whole architectural premise is that your warehouse already contains the modeled, joined, deduped version of your customer data, and the job is to ship that out the door.

2. SQL is the interface. You define what to sync by writing a SQL query (or pointing at a dbt model). The result of that query becomes the rows that get pushed to the destination. This is the killer design choice: it puts data and analytics engineers in the driver's seat, using the language they already speak, rather than requiring Salesforce admin skills.

3. Field mapping and idempotent upserts. The hard part of syncing data to SaaS apps is the tedium: matching warehouse columns to destination fields, handling create-vs-update logic, dealing with rate limits, retrying failed rows. Reverse ETL tools manage all of this so you don't have to.

The Opinionated Take: Reverse ETL Is Now Table Stakes

In 2019, reverse ETL was a venture-backed novel category with two startups. By 2026, it has become table stakes for warehouses themselves. Snowflake, Databricks, and BigQuery have all shipped or partnered for outbound sync features. The Salesforce Data Cloud product — their flagship 2023-2024 launch — is essentially "reverse ETL but Salesforce charges for it." This is what happens to every successful data infrastructure category: the underlying platform absorbs it.

This does not mean Hightouch and Census are dead. Two reasons:

1. The warehouse-native versions are mediocre. They cover the easy destinations (Salesforce, maybe HubSpot) and miss the long tail. Hightouch and Census have hundreds of destinations each, and they invest heavily in the operational details — field mapping, observability, sync history, audit logs — that warehouse vendors don't prioritize.

1. Reverse ETL became the wedge for a bigger product. Both Hightouch and Census expanded upward from "sync warehouse to Salesforce" into adjacent categories: customer data platforms (CDPs), audience builders, marketing campaign orchestration, and identity resolution. The simple sync product is the foothold; the bigger surface is "everything that happens between the warehouse and the marketer." Hightouch in particular has marketed itself as "the warehouse-native CDP" and is competing directly with Segment.

Where Reverse ETL Fits in the Stack

SaaS apps (Salesforce, Stripe, Zendesk)
        ↓ ETL (Fivetran, Airbyte)
Data warehouse (Snowflake, BigQuery, Databricks)
        ↓ dbt models, analytics engineering
Modeled tables (customer_health_score, ltv, churn_risk)
        ↓ Reverse ETL (Hightouch, Census)
Back to SaaS apps (Salesforce, HubSpot, Iterable, Google Ads)

The full loop is sometimes called the "composable CDP" or the "warehouse-native customer data platform": instead of buying a monolithic CDP like Segment that owns ingestion, identity, modeling, and activation, you assemble the same functionality from best-of-breed tools that all share the warehouse as the source of truth. Reverse ETL is the activation half of that loop.

Tools in This Category

• Hightouch — The current leader. Aggressively expanded into customer data platform / activation territory. Strong in marketing use cases.
• Census — The other major player. More analytics-engineering-flavored. Strong dbt integration. Acquired Sutro in 2024.

Polytomic is a smaller third entrant focused on B2B sync use cases. Grouparoo was an open-source player acquired by Airbyte in 2022 and effectively wound down.