RudderStack

Overview

RudderStack is a customer data platform that happens to do reverse ETL, rather than a reverse ETL tool that grew. It began as an alternative to event-collection tools, and event collection is still its centre of gravity, at a published 300 billion events a month.

Like its rivals it stores nothing itself, leaving ownership and control of the data with you, and builds customer profiles inside your own warehouse.

Profiles is the distinctive piece. You declare entities, identity rules and attributes in version-controlled YAML, and it generates the warehouse SQL that builds a customer 360 table and manages the identity graph.

That YAML-first approach is genuinely different from Hightouch's SQL and visual builders and from Fivetran's SQL and dbt route, and it suits engineering teams who want models in a repository.

The licence needs care, and this is where older descriptions go wrong. The server is under the Elastic License 2.0, which is source-available rather than open source: it forbids offering the software as a hosted service. Only the SDKs are genuinely open source, the JavaScript one under MIT.

More importantly for this category, reverse ETL is excluded from the open-source tier altogether, along with models, sync scheduling, single sign-on, private deployment, multi-node scaling and audit logs.

Its consent and governance tooling is the strongest of the three, with client and server-side consent categorisation per destination, tracking plans that flag violations, and a suppression API that deletes a user across destinations.

Its alerting is also the best documented, covering latency, delivery failures, volume drops, transformation failures and partial row failures, across email, Slack, Teams, PagerDuty and webhooks.

It is the only one of the three publishing a flat monthly list price, and the only one with a self-hosted option, though note it has announced no funding since February 2022 while both rivals consolidated or raised heavily.

Key features and capabilities

The same headings are used for every reverse etl & activation entry, so two tools can be read side by side.

Where data comes from

• Nine reverse ETL sources, Snowflake, BigQuery, Databricks, Redshift, Postgres, MySQL, Trino, S3 and SFTP
• Narrower warehouse coverage than the other two, with no Synapse, ClickHouse or Fabric
• Event collection is its core, with open-source SDKs for web, mobile and a dozen server languages
• More than 50 cloud app sources including Shopify, Segment, Braze and Kafka, plus a custom webhook source
• Warehouse destinations double as sinks, so it can both fill and read your warehouse

Business tools it syncs to

• The documentation breaks these down as 300 event stream destinations plus 23 reverse ETL only
• The marketing site says 200 business tools, which contradicts the documentation's own figure
• Four destination types, event stream, warehouse, reverse ETL only, and HTTP webhook
• Connection modes are the published tiering axis, cloud, device and hybrid, hybrid on four destinations
• Per-destination support tiers are not published

Defining audiences and models

• Profiles is declarative, defining entities and attributes in YAML that generates warehouse-native SQL
• Identity stitching rules are declared the same way, with identity graphs managed automatically
• The output is a customer 360 table in your own warehouse
• YAML-first rather than SQL-first or dbt import, which materially differs from both rivals
• Agentic audiences allow natural-language audience definition

Sync behaviour and scheduling

• Three reverse ETL schedule types, a set interval, a cron expression, or manual
• Published intervals from five minutes to 24 hours, with five minutes the documented floor
• Sync frequency is plan-gated, 30 minutes on paid tiers against three hours on free
• Event streaming is real-time; reverse ETL is scheduled batch
• Alerting is the best documented of the three, across email, Slack, Teams, PagerDuty and webhooks

Security, consent and observability

• SOC 2 Type 2, HIPAA with a business associate agreement, plus GDPR and CCPA; no ISO 27001 is published
• The strongest consent management of the three, client and server-side, per destination, with OneTrust and Ketch
• Tracking plans flag unplanned events and bad properties; a suppression API deletes a user across destinations
• Single sign-on, permissions limiting access to personal data, and audit logs, all excluded from open source
• Only SSH tunnelling is published for encryption; no at-rest statement or TLS version appears

Beyond syncing

• A full customer data platform, with event collection, profiles, audiences and in-pipeline transformations
• RudderAI, launched June 2026, adding an MCP server, a command-line tool and agent skills
• The MCP server works with Claude, Cursor and coding agents, with write access deliberately narrow
• Infrastructure as code through Terraform, unique among the three
• No personalisation engine or reinforcement-learning decisioning; agent decisioning is research preview

Where it runs and what it costs

• The only one of the three with a self-hosted option, a Go data plane needing only PostgreSQL
• Self-hosting is governed by the Elastic Licence, and the open-source data plane is single-node
• Multi-node scaling and private deployment are paid-only
• Data residency in the United States or Europe, with routing per geographic requirement
• Billed on events per month

Pricing

SubscriptionFree to 250,000 events a month

The only one of the three publishing a flat list price. Free covers 250,000 events a month forever, with processing stopping after repeated overages. Growth is $265 a month for a million events, scaling through brackets to 25 million, though prices above the first bracket are not published; annual billing saves 15% and a 30-day trial is offered. Enterprise is quote-only. Paid tiers add 30-minute rather than three-hour warehouse syncs and HIPAA. Self-hosting has no licence fee but excludes reverse ETL entirely, which makes it the wrong free option here.

Vendor pricing page →

Demos and videos

• How RudderStack works Product demo · rudderstack.com
• Send data in three minutes Tutorial · rudderstack.com

About RudderStack

RudderStack was founded in 2019 by Soumyadeb Mitra, who remains chief executive; its about page refers to co-founders without naming others, and no headquarters is published on the about, contact or security pages. It is private, having raised a $21m Series A and a $56m Series B in February 2022 led by Insight Ventures with Kleiner Perkins, bringing total funding to $82m, with nothing announced since. It publishes 104 staff and more than 550 customers, including Vercel, Glassdoor, Crate & Barrel and bol.com. The server is Elastic License 2.0, source-available rather than open source; the SDKs are genuinely open.

Founded 2019 · rudderstack.com