Warehouse-native experimentation and feature-flagging — built by ex-Airbnb data scientists, now inside Datadog's product-analytics stack as of May 2025
Eppo is a warehouse-native experimentation and feature-flagging platform that became one of the leading "modern stack" alternatives to Optimizely and LaunchDarkly between 2022 and 2025. The defining design choice: instead of running experiments in its own data silo, Eppo connects directly to your data warehouse (Snowflake, BigQuery, Databricks, Redshift) and runs the statistical analysis where your data already lives. The company was founded in 2021 in San Francisco by Chetan "Che" Sharma, who built Airbnb's experimentation platform before starting it. Eppo raised $47.5M total across Seed, Series A (Menlo Ventures, 2022), and Series B (Innovation Endeavors, 2024), and was acquired by Datadog in May 2025 — reportedly for ~$220M. Eppo continues to be sold as a standalone product but is now also being integrated into Datadog's Product Analytics suite. For Indian product teams: this is the right pick when your data team has already invested in a warehouse and you want experimentation that doesn't fight your existing data model. It is the wrong pick when you're early-stage and need a self-serve A/B tool today.
Eppo is an experimentation, feature-management and product-analytics platform built on a single design principle: your warehouse is the source of truth for everything that matters. Instead of pushing experiment events to a vendor-owned data store and getting a vendor-rendered dashboard back, Eppo's statistical engine runs on top of your warehouse — Snowflake, BigQuery, Databricks or Redshift — and the metrics it computes use your existing fact and dimension tables. That means the same north-star metric your finance team reports on is the metric Eppo uses to call an experiment a winner. No metric drift, no reconciliation arguments.
The company was founded in 2021 by Chetan "Che" Sharma, an early data scientist at Airbnb who was instrumental in standing up Airbnb's experimentation culture and platform. The product reflects that lineage: it bakes in the statistical and metric-design discipline that Airbnb-class teams expect (variance reduction via CUPED, sequential testing, robust treatment-effect estimation, automatic outlier handling). The team raised a Seed, then a $16M Series A in June 2022 led by Menlo Ventures, then a $28M Series B in August 2024 led by Innovation Endeavors — $47.5M total before exit.
On 5 May 2025, Datadog announced the acquisition of Eppo, reportedly for around $220M (per Statsig's analysis of the deal). Eppo continues to be sold standalone and the brand is intact, but the strategic direction is clear: Datadog wants experimentation, feature flags and product analytics in the same pane as its observability stack, and Eppo is the wedge to get there. For Indian buyers, the practical change is in the procurement path — you can now buy Eppo through your Datadog AE if you already have a Datadog contract, which usually shortens the sales cycle.
Connects directly to Snowflake, BigQuery, Databricks or Redshift. The statistical engine runs queries against your existing tables. No event-pushing to a vendor-owned silo, no metric duplication, no reconciliation pain at month-end.
Variance reduction via CUPED, sequential testing, automatic guardrail-metric monitoring, multi-arm bandits, and treatment-effect estimation that is robust to outliers. The most rigorous statistical engine in the commercial experimentation category alongside Statsig.
Server- and client-side feature-flag SDKs in Python, Go, Node, Ruby, Java, Kotlin, Swift, JS. Tied directly to the experimentation surface so flagging a feature and measuring its lift live in one workflow.
Funnels, retention curves, segment cuts — also computed against your warehouse. Less mature than Mixpanel or Amplitude as a standalone product-analytics tool, but the warehouse-native model means metrics line up with Eppo's experiments by default.
Newer surface (post-2024): treat prompt and model variants like any other experiment, measure their impact on user-level metrics. This was a major part of the Datadog acquisition rationale.
Post-acquisition, Eppo experiments and flags are being wired into Datadog's APM, RUM and Product Analytics surfaces. If you already run on Datadog, the integration roadmap is the upside; if you don't, the standalone product still works fine.
Eppo does not publish list prices. Real-world contracts in 2025–2026 typically land in the $30K–$120K+ per year range for the standalone product depending on team size, warehouse data volume, and which surfaces (experiments only vs experiments + flags + product analytics) you license. Post-Datadog there is also a path to bundle inside a Datadog contract, which can move the unit economics either way depending on your existing Datadog spend. For Indian buyers that translates to roughly ₹25L–₹1Cr+ per year all-in. The previous "₹5L+/year" figure on this page (~$6K) understated real list pricing significantly; we've corrected it. Always validate via a quote on geteppo.com.
Eppo is the wrong call when you're an early-stage team that needs an A/B tool live in two days (use VWO or Convert.com), when you don't have a warehouse (use GrowthBook open-source), or when feature flags are your only need (use LaunchDarkly).