Skip to the content
Know your tech before you buy it. Makers: get listed from $49/mo

LaunchDarkly Experimentation

launchdarkly.com

  • Free plan
  • Web
  • iPhone and iPad
  • Android
  • API
7.7 Club score Visit launchdarkly.com

What LaunchDarkly Experimentation is

LaunchDarkly Experimentation helps teams evaluate changes using behavioral results, with tests covering features, models, prompts, page layouts, and AI-powered systems. Teams can segment results by cohort, device, geography, or attribute, then export them to a warehouse for further analysis. Statistical tools include CUPED variance reduction, sequential testing, sample ratio mismatch detection, and multi-armed bandits. Mutual exclusion and holdout groups can isolate experiments and help measure longer-term impact. Warehouse-native experimentation supports Snowflake, Redshift, and Databricks, among other systems. Listed integrations also include Amplitude, Datadog, Slack, and Microsoft Teams. Client-side and mobile SDKs include JavaScript, iOS, and Android, with a separate API for programmatic product interactions. The free Developer plan includes 100K experimentation MAU/month, one project, three environments, and five service connections/month. Foundation is $108.33 a month billed yearly, with a 14-day trial; paid plans start at $8.33/mo (annual). LaunchDarkly lists SOC 2 Type II, ISO 27001, FedRAMP authorization, and Data Privacy Framework participation.

Who it suits

It suits product, marketing, and business teams as well as developers, including product managers and data scientists. Teams that want to analyze experiments by cohort or export results to a supported warehouse may find its listed tools relevant.

Good to know, and what to watch

  • Tests features, models, prompts, layouts, and AI systems.
  • Segments results by cohort, device, geography, or attribute.
  • Includes holdout groups and mutual exclusion controls.
  • Free Developer plan includes 100K experimentation MAU/month.
  • Mobile SDKs include iOS and Android.
  • Developer plan is limited to one project.
  • Developer plan includes five service connections/month.
  • Foundation is $108.33 a month billed yearly.
  • Enterprise pricing is custom and not listed.

Enterprise

—

Developer

Free

Foundation

$108.33/mo

Free plan · 14-day trial

The Club's review

Overview

LaunchDarkly Experimentation is an experimentation service for measuring how product changes affect user behavior across web and mobile. It is best suited to product and data teams that need controlled tests, statistical methods and warehouse analysis; its breadth is a strength, but the free plan’s monthly connection cap and the annual billing for Foundation are meaningful constraints.

Teams can experiment on features, models, prompts, page layouts and AI-powered systems. The service also includes feature flags, audience targeting and server-side testing, making it a better fit for product development workflows than for teams seeking a visual page-testing editor.

Key features

Experiment design and controls

Experiment types include A/B and A/A tests, multi-armed bandits and holdout experiments. CUPED variance reduction, sequential testing and sample ratio mismatch detection add useful statistical tools for teams that need to assess results carefully. Mutual exclusion helps isolate experiments from one another, while holdout groups let teams examine longer-term impact. This range is useful beyond straightforward conversion tests, especially for teams evaluating models or AI prompts.

Analysis and data connections

Results can be segmented by cohort, device, geography or attribute, then exported to a warehouse for further analysis. Warehouse-native experimentation supports Snowflake, Redshift and Databricks, among others; listed integrations also include Amplitude, Datadog, Google Cloud Pub/Sub, Slack and Microsoft Teams. Client-side and mobile SDKs include JavaScript, iOS and Android, and a separate API supports programmatic product interactions.

The service has no visual editor, so teams looking to build page experiments through a visual interface should look elsewhere. Security provisions include SOC 2 Type II, ISO 27001 and FedRAMP authorization, along with encryption in transit, MFA, SAML SSO and audit logging. LaunchDarkly says it does not require personally identifiable information and offers an EU-hosted region for EU data residency.

Pricing

The Developer plan is free forever at no cost. It includes 100K experimentation MAU/month, 1 project, 3 environments, 5 service connections/month, and A/B tests and experiments. That is a practical starting point for a small team, but the project and connection caps limit how far it can stretch across products or integrations.

Foundation costs $108.33 a month, billed yearly, with a 14-day trial followed by pay-as-you-go billing. It includes the first 5 service connections/month, 5K AI runs/month, and unlimited projects and environments; additional runs cost $5 per 1k. The extra capacity suits teams that have outgrown Developer, though the annual billing term is a substantial commitment and AI-run overages should be part of the budget.

Enterprise has custom pricing and includes everything in Foundation, a custom AI run count, 100 days of data retention, advanced targeting, and roles, teams, workflows and approvals. It is aimed at organizations needing more governance and tailored AI capacity rather than teams seeking a published fixed price.

Platforms

LaunchDarkly supports Android, iOS, web and API access. Mobile SDKs and server-side testing make it relevant to teams testing changes across app and service environments, not just browser experiences.

Who it's for

Product managers, data scientists, developers, and product, marketing or business teams can use the service to measure changes and analyze outcomes. It is most compelling where experiments need isolation, statistical rigor and warehouse workflows. A small team running occasional page tests may find its broader platform and lack of a visual editor unnecessary.

Pros and cons

  • Pros: Multiple test types and methods, including CUPED, sequential testing and bandits, support more sophisticated experimentation than basic A/B testing alone.
  • Pros: Segmentation and warehouse export connect experiment results to broader data analysis, with Snowflake, Redshift and Databricks support.
  • Pros: Mutual exclusion and holdout groups help teams reduce experiment interference and evaluate longer-term effects.
  • Cons: Developer is limited to 1 project and 5 service connections/month, which can constrain teams with several products or data connections.
  • Cons: Foundation is billed yearly, and AI runs beyond the included 5K/month cost $5 per 1k.
  • Cons: There is no visual editor for teams that prefer to create page tests without SDK or product-development workflows.

Alternatives

Statsig is worth considering for teams that want a free Developer tier with 2M events/month, unlimited flag and config checks, 50,000 session replays/month and one-year analytics retention; its Pro plan is $150 a month.

ConfigCat may suit teams focused on feature flags, with a Forever Free plan covering 10 flags, 2 environments and 5 million config JSON downloads/month. Flagsmith is another option for teams that prioritize open-source feature flag functionality and want to consider its supported self-hosted and desktop platforms.

Devtron Release Orchestration, ReARM, Runway, Blue Canvas and Orangescrum are alternatives to compare.

For broader category browsing, see A/B Testing Tools, Feature Flag Software and Release Management Software.

Verdict

LaunchDarkly Experimentation is a strong choice for product and data teams that need rigorous experiments connected to feature delivery and warehouse analysis. Its statistical breadth and controls are the main reasons to choose it; teams needing a visual editor, or unwilling to take on Foundation’s annual billing and usage charges, should look elsewhere.

The verdict

LaunchDarkly Experimentation combines experiment controls, statistical methods, and warehouse-oriented analysis across web and mobile SDKs. The free plan has defined project and connection limits, while Foundation pricing is listed annually and Enterprise pricing is custom.

What the maker publishes

Analysis
Results can be segmented by cohort, device, geography, or attribute, and exported to a warehouse for further analysis. source
Company
LaunchDarkly was founded in 2014 in Oakland, California, by Edith Harbaugh and John Kodumal. source
Experiment controls
Users can use mutual exclusion and holdout groups to isolate experiments and measure long-term impact. source
Experiment scope
Teams can test features, models, prompts, page layouts, and AI-powered systems. source
Founded
2014 source
Headquarters
Oakland, California, United States source
Integrations
Listed integrations include Databricks, Snowflake, Redshift, Amplitude, Datadog, Google Cloud Pub/Sub, Slack, and Microsoft Teams. source
Privacy and residency
LaunchDarkly says it does not require personally identifiable information and offers an EU hosted region supporting EU data residency. source
Purpose
Experimentation helps teams measure impact, learn from real behavior, and improve what they ship. source
SDKs and API
LaunchDarkly lists client-side and mobile SDKs including JavaScript, iOS, and Android, and documents a separate API for programmatic product interactions. source
Security
The company lists SOC 2 Type II, ISO 27001, FedRAMP authorization, and Data Privacy Framework participation, along with encryption in transit, MFA, SAML SSO, and audit logging. source
Statistical methods
Experimentation includes CUPED variance reduction, sequential testing, sample ratio mismatch detection, and multi-armed bandits. source
Support
The site links to a Help Center and a direct support request channel. source
Warehouse support
Warehouse-native experimentation supports Snowflake, Redshift, and Databricks, among others. source
Who it serves
LaunchDarkly says its platform serves product, marketing, and business teams as well as developers, and names product managers and data scientists among its target teams. source

Checked against the maker's pages on 29 Sept 2026.