GrowthBook

The top one percent of companies spend thousands of hours building experimentation, feature flagging and analytics in house. GrowthBook exists so the other 99 percent get the same platform without paying for expensive third party SaaS or hacking together unmaintained libraries.

It is warehouse native, querying eleven data sources including BigQuery, Snowflake and Databricks rather than keeping its own copy of your events. Metrics are defined in SQL, from simple conversion rates to ratios and quantiles, and 24 SDKs cover React, Python, Kotlin, Swift and more.

The platform pairs the rollout side with the measurement side.

• Targeted flags: support advanced targeting, gradual rollouts and experiments from a single definition.
• Stats engine: runs CUPED, sequential and Bayesian analysis, post-stratification, bandits and SRM checks.
• Analytics suite: builds dashboards from warehouse data and shares them with the team.
• Documentation in place: attaches screenshots, custom meta fields and Markdown throughout the app.
• Webhooks and REST API: drive integrations and custom workflows.
• MCP server: creates features, starts experiments and cleans up stale flags from an agent.

GrowthBook is open core, most of the code under the permissive MIT license with several directories governed by a separate commercial enterprise license. Self-hosting is a clone and a docker compose up, using the prebuilt public image that needs no registry credentials, with the app waiting on port 3000.