Eppo, now Datadog Experiments, is a warehouse-native platform for A/B testing, feature flags, and statistical experiment analysis used by Coinbase and Twitch.
Eppo started in 2021 as a standalone experimentation and feature-management platform built around the idea that A/B test data should live and be analyzed directly inside a company's own data warehouse rather than in a separate vendor silo. Instead of exporting events to a proprietary database, Eppo connects to Snowflake, BigQuery, Redshift, or Databricks and runs its statistical engine against that existing data, which keeps sensitive information in place and lets analytics teams reuse metrics they already trust.
In May 2025 Datadog acquired Eppo in a deal reported at roughly 220 million dollars, folding the product into its observability suite. The tool is now sold as Datadog Experiments, though the original Eppo brand and dashboard experience continue for existing customers while Datadog integrates it with its broader monitoring, feature-flag, and AI-evaluation tooling.
Eppo's core is a statistics engine that supports fixed-sample, sequential, and Bayesian analysis methods alongside CUPED and CUPED++ variance reduction, which lets teams reach statistically significant conclusions faster with less traffic. Built-in metric governance lets analytics and data science teams define a shared library of vetted metrics so that every team running an experiment measures success the same way, reducing the common problem of different teams reporting conflicting results from the same test.
Beyond pure A/B testing, Eppo bundles feature flagging and progressive rollout controls so engineers can gate new code behind flags and gradually expose it to more users while the experimentation engine measures impact in real time. It also supports use cases like AI model evaluation and lifecycle marketing experiments, and connects to common event pipelines and BI tools so results can be surfaced where teams already work.
Eppo does not publish self-serve pricing. Since its launch it has followed an enterprise sales model where prospective customers request a demo and receive a quote based on data volume, number of experiments, and seats, and that pattern has continued under Datadog ownership.
Because the tool now sits inside Datadog's commercial ecosystem, pricing conversations increasingly happen alongside a company's broader Datadog contract, and existing Datadog customers may find it bundled or discounted as part of a larger observability and experimentation package rather than sold as a fully separate line item.
Yes, existing and new customers can still use the Eppo platform, though it is now marketed under the Datadog Experiments name following Datadog's 2025 acquisition of the company.
Eppo does not list public pricing. Costs are quoted per deal based on data volume, seats, and experiment scope, and are typically discussed through a sales demo.
It is built to connect to popular cloud warehouses including Snowflake, BigQuery, Redshift, and Databricks; a warehouse is effectively a prerequisite for getting full value from the platform.
Yes, Eppo includes native feature flagging and progressive rollout controls alongside its experimentation engine, so many teams use it instead of a standalone flagging service.
Data scientists and analytics engineers usually own metric definitions and experiment design, while product managers and engineers use the flagging and rollout tools day to day.