Eppo

Eppo provides an end-to-end experimentation platform combining A/B testing analysis, feature flagging, personalization and AI model evaluation. Its architecture is warehouse-native: experiment analysis runs against a customer's existing data warehouse rather than copying data into a vendor system, which the company positions as reducing data egress, avoiding conflicting data sources, and controlling warehouse compute costs. Metric definitions, guardrails, experiment protocols and workflows are centralized and version-controlled through a semantic layer so that data teams retain ownership of definitions across the organization.

The statistical engine supports sequential, fixed-sample and Bayesian analysis frameworks, along with CUPED variance reduction (marketed as CUPED++) intended to shorten experiment runtimes. Alongside analysis, Eppo offers feature flags used for A/B tests, feature gates, controlled rollouts, kill switches and dynamic no-code configuration, which the company says power billions of daily assignments, as well as contextual bandits for real-time personalization and tooling to evaluate AI models against business metrics. Additional capabilities include automated deep dives and diagnostics, self-service slice-and-dice analysis for product managers, and Experiment Forecasts for roadmap planning.

Eppo has been acquired by Datadog and the product is presented as Datadog Experiments.

Business model

Eppo sells a software platform to companies running product experiments, layering white-glove support and access to experimentation practitioners alongside the product. Following the acquisition, the offering is delivered as part of Datadog.

Latest developments

Eppo has been acquired by Datadog and is being offered as Datadog Experiments, with the company describing a year spent building a new generation of self-service experimentation and feature flagging tooling now available on Datadog.