CRO and testing · guide

GrowthBook: feature flags and A/B tests

GrowthBook is an open-source feature flag and experimentation platform. It analyses tests directly on the data in your warehouse, so results are measured with your own core metrics. It is one of the most used alternatives since Google Optimize closed.

Type
Feature flags and A/B tests
Price
Open source, Cloud plans
Analysis
Warehouse-native
Statistics
Bayesian and frequentist

What is GrowthBook?

A GrowthBook test has two parts: the SDK splits users into variants, and the analysis runs on event data in BigQuery, PostHog or another source. There is no need for a separate tracking system.

Feature flags are also risk management: you release a new feature to a small share first and switch it off without a code change if something goes wrong.

What GrowthBook makes possible

  1. A/B and multivariate tests

    Statistically sound results.

  2. Feature flags

    Gradual roll-out and a kill switch.

  3. Visual editor

    Simple web tests without code.

  4. Warehouse-native analysis

    With your own metrics and data.

  5. Segment results

    By device, source and country.

  6. Guardrail metrics

    Making sure a test does not hurt another metric.

What GrowthBook works with

  • Google BigQuery

    A serverless data warehouse. GA4 raw data and large-scale SQL analysis.

    Open the guide
  • PostHog

    Product analytics, session recordings, feature flags and tests in one place.

    Open the guide
  • Google Analytics 4

    Event-based web and app analytics. Funnels, conversions and audiences.

    Open the guide
  • GitHub

    Code hosting, version control, teamwork and CI/CD.

    Open the guide

My experience

Projects where I used GrowthBook:

  • [Project: brand, what I built, result %]
  • [Project: brand, what I built, result %]
See the cases

GrowthBook resources

  • GA4 event plan

  • CRO service: who it is for

  • Microsoft Clarity guide

Questions about GrowthBook

Is GrowthBook free?

The open-source version is free. The Cloud version has free and paid plans.

How much traffic does a test need?

It depends on the baseline conversion and the expected effect. I calculate the sample size before the test.

Do I need a developer?

For the SDK setup and complex tests, yes. Simple visual tests can be run by the marketing team.

Other tool guides