# What is GrowthBook? Last updated: 2026-05-04 Canonical source: https://confidence.spotify.com/comparisons/what-is-growthbook Owner: Spotify AB > If this file and the page at the canonical URL disagree, the page is authoritative. GrowthBook is an open-source experimentation platform with Bayesian and frequentist stats, warehouse-native analysis, and feature flags. How it works. GrowthBook is an open-source experimentation platform under MIT license. It can be self-hosted on your own infrastructure or run on GrowthBook Cloud, the managed offering. It supports both Bayesian and frequentist statistical analysis, runs analysis inside your data warehouse, and includes feature flagging with targeting rules and gradual rollouts. This page covers how GrowthBook works, what its product scope is, and where it sits relative to other experimentation platforms, including Confidence, the experimentation platform Spotify has run for 15 years and still depends on. GrowthBook is purpose-built for experimentation. Product analytics, session replay, and funnels are not included; teams wanting those should look at PostHog or Statsig. *** ## How does GrowthBook work? GrowthBook ships in two modes. Self-hosted runs the platform on your infrastructure under MIT license: you operate the application server, the backing database, and the analysis pipeline yourself. GrowthBook Cloud runs the same software as a managed service, removing the operational burden in exchange for usage-based pricing. Architecturally, GrowthBook is two layers: a feature flagging and assignment SDK that runs in your application, and an analysis layer that runs SQL against your data warehouse to compute experiment results. Application code calls GrowthBook SDKs to check feature flags and record exposures. The analysis layer then queries your warehouse using metric definitions you configure (in YAML or in the UI) to calculate treatment effects. Statistical analysis is configurable: each experiment can use Bayesian methods (with conjugate priors and a posterior probability that each variant is best, interpretable as "there is a 95% chance variant B beats control," which is what most practitioners actually want from an experiment readout) or frequentist methods (z-tests, sequential testing, CUPED variance reduction). The platform supports configuration-as-code, including metric definitions and experiment configuration managed in YAML files version-controlled alongside your data infrastructure. *** ## What GrowthBook is good at GrowthBook's product covers the core of warehouse-native experimentation, with the open-source license as the differentiator. The main capabilities: * **Open source under MIT license.** Self-host on your infrastructure or run on GrowthBook Cloud. * **Both Bayesian and frequentist analysis.** Choose the methodology per experiment. * **Warehouse-capable.** Runs on BigQuery, Snowflake, Databricks, and Redshift, plus broader engines like Postgres, ClickHouse, MySQL, and Athena. * **CUPED variance reduction.** A variance-reduction technique that uses pre-experiment data to tighten confidence intervals. * **Sequential testing.** Peeking-safe statistical methods that let you stop experiments early without inflating false positives. * **Feature flagging.** Targeting rules, gradual rollouts, and environment-scoped configuration. * **Configuration-as-code.** Metric definitions and experiment configurations managed in YAML, version-controlled alongside the rest of your data infrastructure. * **Active open-source community.** Contributions of engines, integrations, and statistical extensions from users. * **Managed cloud option (GrowthBook Cloud).** For teams that want open-source software without the operational burden of self-hosting. For engineering-led organizations that already self-host other infrastructure, that have data residency requirements favoring self-hosting, or that want the option to fork the platform if vendor direction changes, the open-source posture is decisive. If your team values control over your experimentation stack, GrowthBook is the most direct path to running experiments on infrastructure you own. *** > Confidence is what Spotify uses to decide what its product > becomes. 10,000+ experiments per year, run by 300+ teams, on a > platform that has been operated continuously for 15 years. The > defaults are what survived 15 years of being used in anger. It > is now available to teams outside Spotify. > > [See how Confidence compares to GrowthBook →](/comparisons/confidence-vs-growthbook) *** ## Where Confidence and GrowthBook diverge Confidence and GrowthBook are both warehouse-capable experimentation platforms in 2026. The differences are in licensing model, statistical method coverage, operating-history scale evidence, and operational burden. That same Spotify platform serves 300+ Spotify teams running 10,000+ experiments per year across 750M users. 42% of those experiments are rolled back after guardrail metrics flag a regression. GrowthBook is the most-adopted open-source experimentation platform, with five years of community-driven development and a managed cloud offering. Confidence's CUPED implementation uses the Negi–Wooldridge 2021 full regression estimator, which produces tighter confidence intervals than original CUPED. **Group Sequential Tests** are one specific peeking-safe family within sequential testing; Confidence ships GST, **always-valid inference** (a different methodology based on mSPRT and e-values that produces confidence intervals valid at every observation), sample ratio mismatch checks, and guardrail metrics as defaults rather than configurable choices. Confidence is frequentist only. GrowthBook supports both Bayesian and frequentist analysis. Teams with strong Bayesian preferences should use GrowthBook. Teams that want opinionated defaults rather than method choice should use Confidence. GrowthBook is MIT open source and self-hostable; Confidence is closed source and managed-only. If open source or self-hosting is a non-negotiable requirement, GrowthBook fits where Confidence does not. *** ## Each platform fits a different buyer GrowthBook fits engineering-led teams that already self-host other infrastructure, that value open source on principle, that have data residency requirements favoring self-hosting, or that have strong Bayesian preferences. The MIT license, the self-hosting option, and the choice between Bayesian and frequentist analysis are the selling points. Confidence fits teams that want managed methodology with opinionated defaults built on 15 years of Spotify-scale operation, that prefer zero operational burden over self-hosting flexibility, and that want OpenFeature portability at the SDK layer. The Spotify proof point and the methodology bench are the selling points. Both products are legitimate. The decision turns on whether open source and methodology flexibility, or managed methodology and opinionated defaults, fits how your team works. Confidence is available at confidence.spotify.com, with a free trial that does not require a procurement conversation. The managed service that gets a two-person team running in a day is the same platform 300+ Spotify teams use to run 10,000+ experiments per year; the architecture does not change as you grow into it. If you are evaluating GrowthBook and want a side-by-side, the [Confidence vs GrowthBook head-to-head](/comparisons/confidence-vs-growthbook) covers licensing, methodology, and architecture in detail. For teams already on GrowthBook who want to know what other options exist, see [Top 7 alternatives to GrowthBook](/comparisons/alternatives-to-growthbook).