Eppo review

What happened to Eppo

Eppo was a San Francisco experimentation company led by co-founder and CEO Che Sharma. Datadog announced it had acquired Eppo in May 2025. Terms weren't disclosed, and TechCrunch cited a report putting the deal at about $220M. At first it ran as Eppo by Datadog. On April 2, 2026 Datadog launched Datadog Experiments, powered by Eppo, as a generally available product, and geteppo.com now says Eppo is Datadog Experiments. The Eppo docs site is still online under the Eppo by Datadog name.

How it works

Eppo sits on top of your data warehouse and supports Snowflake, BigQuery, Redshift and Databricks. You map exposure records and business metrics to it with SQL models, so results come from the same tables your data team already uses. In Datadog you can also build metrics from Product Analytics or RUM events, such as a count of events, unique users or the sum of a property.

Assignment runs through Datadog Feature Flags, where you pass a stable user ID as the targeting key, or through your own randomization in warehouse mode. You don't need Product Analytics to run experiments.

There's no point-and-click page editor. Eppo's docs describe no-code tests through JSON flags, where developers parameterize parts of a page once and then change values in Eppo without a release. A Contentful integration lets marketers create experiments on CMS content inside Contentful.

Stats

Datadog Experiments offers sequential frequentist, fixed-sample frequentist and Bayesian analysis. It adds CUPED++ variance reduction, a sample size calculator, automatic sample ratio mismatch detection and a Global Lift view that turns a targeted result into company-wide impact.

Inside Datadog

The Datadog version connects experiments to observability. If you use RUM, you can watch latency, errors and Core Web Vitals per variant to catch performance regressions, and open session replays from the results.

Revenue

There's no payment tool connector. Revenue metrics work if your Stripe or billing data already lands in your warehouse. If it doesn't, you need a pipeline first.

AI

Neither the Datadog Experiments product page nor its docs describe AI that suggests tests or writes variations.

Who it suits

It suits companies with a data warehouse, a data team and a few high-value experiments a month, especially if they already run Datadog. Customers named on geteppo.com include DraftKings, ClickUp, Zapier, Miro and Twitch. It's a poor fit for a small team that wants to test landing pages this week without writing SQL.