Datadog Experiments

Founded 2021

Run data warehouse-native tests and feature rollouts with advanced statistical analysis

You can use Eppo to run reliable A/B tests and feature flags directly on your data warehouse. It connects to your existing data infrastructure without copying data, letting teams evaluate experiments against core business metrics instead of vanity measurements. You can safely roll out new product features, automate test diagnostics, evaluate machine learning models, and personalize user experiences using contextual bandits. The system supports frequentist, Bayesian, and CUPED statistical methods to deliver fast, accurate results while keeping your cloud warehouse compute expenses down.

What Datadog Experiments does

Set up targeted feature flags for staged deploymentsCalculate statistical significance for experiment metricsDefine guardrail metrics to monitor unintended feature side effectsPerform slice-and-dice data analysis on experiment variationsTrigger automated kill switches when performance anomalies occurEvaluate model performance before full production rolloutConfigure contextual bandit rules for automated content selectionRuns A/B test calculations directly inside your data warehouseSupports frequentist, Bayesian, and CUPED variance reduction statisticsProvides feature flagging with controlled rollouts and kill switchesUses contextual bandits for dynamic real-time personalizationEvaluates machine learning and AI models against business metricsGoverns metric definitions and guardrails using a semantic layerEliminates data duplication by keeping data within existing warehouses