GrowthBook

Visit Website

Introduction

GrowthBook describes itself as open-source feature flags, experimentation and product analytics. It combines a flag service with an experimentation platform, so the same flag that controls a rollout can run an A/B test and report its impact.

Its distinctive design is warehouse-native analysis. Instead of collecting all events itself, GrowthBook can query experiment exposure and metric data already in your data warehouse, such as BigQuery, Snowflake, Redshift, Databricks, ClickHouse or Postgres, and compute results there. Teams keep a single source of truth for metrics and avoid copying data to another vendor. GrowthBook Cloud also offers managed event storage for teams without a warehouse.

The stats engine supports Bayesian and frequentist methods, sequential testing to allow early peeking with controlled error rates, CUPED variance reduction, guardrail metrics and multiple-testing corrections. Reports show uplift, confidence and data-quality checks such as sample ratio mismatch, which catches broken assignment before anyone trusts a result.

SDKs are lightweight and evaluate flags locally, with libraries for JavaScript, React, Node.js, Python, Go, Ruby, PHP, Java, Kotlin, Swift, Flutter and more. A visual editor allows simple front-end experiments without code, and a URL redirect test type handles page-level tests.

Most of the code is MIT licensed, with an enterprise directory under a separate licence. The free self-hosted and cloud tiers cover core flags and experiments; paid plans add features such as advanced permissions and enterprise analysis.

In the quality loop GrowthBook straddles production and UX validation: it gates releases with flags and measures whether a change actually improved user outcomes.

The source code is at growthbook/growthbook on GitHub under the MIT (with an enterprise directory under a separate licence) licence, with about 8,600 stars; the most recent commit was on 11 November 2026.

Before trusting experiment results, run an A/A test to confirm that assignment and metrics behave as expected, and decide primary metrics and minimum run time before launch. Treat feature flags created for experiments as temporary and remove them after rollout to avoid flag debt.

Features

1. Feature flags: Targeting, rollouts and environments.
2. A/B testing: Experiments tied directly to flags.
3. Warehouse-native: Query metrics in your own warehouse.
4. Bayesian and frequentist: Choice of statistics engine.
5. Sequential testing: Valid results with early looks.
6. CUPED: Variance reduction for faster results.
7. Data quality checks: Sample ratio mismatch and other warnings.
8. Visual editor: No-code front-end experiments.
9. Lightweight SDKs: Local evaluation across many platforms.
10. Product analytics: Dashboards and metric exploration.

Pros

1. Experiment rigour: Modern statistics features usually found in in-house platforms.
2. Data ownership: Analysis can stay inside your warehouse.
3. Free to use: Open source under the MIT (with an enterprise directory under a separate licence) licence, so the core tool costs nothing to use.
4. Actively maintained: The repository had commits as recent as 11 November 2026.
5. Community: About 8,600 GitHub stars.

Cons

1. Warehouse skills needed: Metric definitions require SQL and data modelling.
2. Mixed licence: Enterprise directory is not under MIT.
3. Setup effort: Self-hosting needs MongoDB and configuration.
4. Experiment design still matters: Tools do not fix underpowered tests.

Usage and Links (Example)

Run self-hosted with Docker:

git clone https://github.com/growthbook/growthbook.git
cd growthbook && docker compose up -d

Evaluate a flag in JavaScript:

import { GrowthBook } from "@growthbook/growthbook";
const gb = new GrowthBook({
  apiHost: "https://cdn.growthbook.io",
  clientKey: process.env.GB_CLIENT_KEY,
  attributes: { id: user.id, country: user.country },
  trackingCallback: (experiment, result) => analytics.track("Experiment Viewed", {
    experimentId: experiment.key, variationId: result.key }),
});
await gb.init();
if (gb.isOn("new-onboarding")) showNewOnboarding();

Practical tips:

1. Define guardrail metrics, such as errors and latency, for every experiment.
2. Run an A/A test first to validate assignment and tracking.
3. Use sequential testing if stakeholders will check results early.

Resources:

• GrowthBook website
• GitHub repository

Pricing

GrowthBook is free to self-host (open source) and has a free cloud tier with no credit card. The Pro plan is listed at $40 per seat per month; cloud usage beyond included allowances is metered (for example 2M events per month, then $30 per million). Enterprise pricing is via sales.

Recommended for?

1. Product teams running regular A/B tests.
2. Data teams with a warehouse who want experiment analysis on their own data.
3. Engineering teams wanting flags and experiments in one tool.
4. Not ideal for teams with no analytics data yet; start by instrumenting events.

Conclusion

GrowthBook brings serious experimentation statistics and feature flags together in an open-source package. Warehouse-native analysis keeps metrics consistent with the rest of the business.

It rewards teams that already have good event data and some SQL skills; with those in place, it is one of the strongest open options for testing changes on real users.

On CheckOps, GrowthBook is filed under the Release & CI gates, Production monitoring & observability and Prototype & UX testing stages. Related tools listed here include Statsig, PostHog, Optimizely, Wingify (VWO and AB Tasty), LaunchDarkly and Microsoft Clarity. Which one fits best depends mostly on your stack, your team’s size and skills, and whether you prefer to run tools yourself or pay for a managed service. This page reflects the official website, documentation and repository as they stood on 11 November 2026.