Hightouch

Hightouch is the activation half of the “composable CDP” argument: instead of a vendor pipeline building profiles in its own store, your warehouse holds the truth and Hightouch syncs it outward. The mental model is sources → models → syncs. A source is your warehouse (Snowflake, BigQuery, Redshift, Databricks, Postgres, and around thirty others, plus oddballs like Google Sheets, S3, Kafka, and Looker Looks). A model shapes that source into a dataset — written as SQL, picked with a visual table selector, or imported directly from dbt or Looker. A sync maps model columns to a destination’s API fields and pushes changes on a schedule.

The sync layer is where the engineering substance lives. Sync modes are explicit: upsert, update, insert, mirror (“all”), plus snapshot and diff modes for file destinations and add/remove for list membership. Change detection means only new or changed rows are pushed on each run rather than the full result set — event-style syncs should use insert mode with a unique event ID column so nothing gets deduplicated away. Scheduling runs from the UI, chained Sequences, or triggers from an orchestrator (Airflow, Dagster, Prefect via the API), which matters if syncs must fire only after dbt models finish building.

Implementation notes worth knowing before you commit. First, the “no stored copy” pitch has grown qualifiers: Hightouch now sells Events (SDKs for browser, iOS, Android, and Node, an HTTP API, and Kafka/Pub-Sub connectors) that lands event data in your warehouse and can stream to destinations, plus an Identity Resolution product that builds unified profiles on warehouse tables — so the platform can now cover collection and identity too, but each is a separately priced component, not something you get with reverse ETL. Second, Customer Studio adds a visual audience builder, journeys, and measurement on top of warehouse data so marketers can segment without SQL — again priced separately. Third, Match Booster enriches audience syncs with additional identifiers to raise ad-platform match rates, which is genuinely useful for conversions/audience use cases but is another line item. Budget conversations should assume the composable pieces add up.

Pricing is usage-based with no MTU metric and no caps on sources, destinations, or seats; the free plan is real but small — two active syncs. Everything beyond that is quoted, so expect a sales cycle rather than a published price list.

Use it when you have a modeled warehouse (especially with dbt discipline) and want CRM fields, audiences, and ad-platform conversions fed from it instead of from a tag. Look past it when you have no warehouse practice yet — every capability sits on top of your tables — or when you want one bundled price for collection, identity, and activation rather than assembling and paying for components individually.