Why Datadog Acquired Eppo

by DataMarvin

9 hours ago

In May 2025, Datadog completed the acquisition of Eppo, an experimentation and feature management platform. The deal price was not officially disclosed, but was widely reported at around $220 million. The legal close happened quickly — but the product integration is still in progress. As of writing, Eppo continues to operate under the "Eppo by Datadog" brand as a separate product while the two platforms work toward deeper integration.

This post focuses on the strategic reasoning behind the deal — why Datadog wanted Eppo, and what it signals for the experimentation market.

1. What Is Eppo?

Eppo is an experimentation and feature management platform built around a specific architectural philosophy: connect directly to your existing data warehouse, and use a rigorous statistical engine to measure how product changes affect business metrics.

Unlike some experimentation tools that store and process data in their own silos, Eppo was designed to sit on top of infrastructure teams already have — Snowflake, BigQuery, Databricks — and apply statistical analysis there. The result is a platform that feels closer to a data team's workflow than a typical A/B testing SaaS.

Its statistical engine supported advanced methods: sequential testing, CUPED-style variance reduction, and guardrail metrics — the kind of experimentation infrastructure that used to exist only at companies like Airbnb, Netflix, or Lyft.

2. What Is Datadog?

Datadog is an observability and monitoring platform for cloud applications. It tracks infrastructure health, application performance, logs, security events, and user experience — giving engineering teams a unified view of what's happening inside their systems.

Until this acquisition, Datadog's focus was primarily on the engineering layer: is the service up? is the latency acceptable? is there an error spike? It had expanded into Product Analytics and Real User Monitoring — but the missing piece was a systematic way to measure the business impact of changes, not just their technical behavior.

3. The Gap Datadog Was Filling

Here's the problem the acquisition solves.

A typical software deployment looks like this:

Engineer ships a change
        ↓
Datadog monitors: latency, error rate, infrastructure health  ← Datadog's strength
        ↓
Did user behavior actually improve?      ← gap
Did revenue go up?                                  ← gap
Which variant performed better?            ← gap

Datadog could tell you if the deployment broke something. It could not tell you if it worked — in the sense of moving the metrics that matter to the business.

Changes were often rolled out without understanding their impact on KPIs, making it difficult to tie those changes back to business outcomes. Experimentation infrastructure — the kind Eppo provides — is exactly the layer needed to close that gap.

4. Why Eppo Specifically?

Several experimentation platforms exist: Optimizely, LaunchDarkly, Statsig, GrowthBook, among others. What made Eppo the acquisition target?

Data warehouse-native architecture. Eppo's design philosophy — connecting directly to the customer's existing data warehouse rather than ingesting data into its own store — aligns well with how modern data teams operate. It reduces data duplication and fits naturally into existing analytics stacks.

Statistical rigor. Eppo invested heavily in the statistical engine: sequential testing, variance reduction, and proper multiple testing corrections. This is harder to build than it looks, and it's the part of experimentation infrastructure that most tools compromise on.

AI experimentation angle. The use of multiple AI models increases the complexity of deploying applications in production. Experimentation solves this correlation and measurement problem, enabling teams to compare multiple models side-by-side, determine user engagement against cost tradeoffs, and ultimately build AI products that deliver measurable value. As AI-powered features become standard, the ability to run rigorous experiments on LLM outputs, prompts, and model versions becomes a first-class need — not an afterthought.

5. What Datadog Gets

The vision is to bring end-to-end observability, feature management, and experimentation into a single, unified platform. The plan is to integrate Eppo's capabilities into Datadog Product Analytics, Real User Monitoring, and Session Replay — though as of now, Eppo still operates separately under the "Eppo by Datadog" brand, with deeper product integration still ongoing.

When fully integrated, the combined platform is intended to cover the full development feedback loop:

Ship a change (feature flag)
        ↓
Monitor technical health (Datadog observability)
        ↓
Measure user behavior (Real User Monitoring, Session Replay)
        ↓
Quantify business impact (Eppo experimentation engine)
        ↓
Decide: roll out, roll back, or iterate

This is the loop that companies like Facebook, Airbnb, and Netflix built internally over years. Datadog is packaging it as a commercial platform.

6. The Broader Signal: Consolidation in Experimentation

The acquisition is also a signal about where the experimentation market is heading.

This acquisition reinforces two fundamental beliefs about the experimentation market: experimentation is central to the modern development stack, and point solutions are being consolidated into a single product development platform.

For years, experimentation infrastructure was assembled from separate tools: a feature flag service here, a stats engine there, a data pipeline in between, a dashboard on top. Each piece was best-in-class in isolation but painful to stitch together. The trend — accelerated by this acquisition — is toward integrated platforms where observability, feature management, and experimentation share the same data layer.

At $220 million, Datadog is making a clear statement: the ability to measure what you ship is as important as the ability to monitor how it runs.

7. What This Means for Teams Using GrowthBook, Statsig, or Similar Tools

The acquisition doesn't make other experimentation platforms obsolete — but it changes the competitive landscape.

The main shift: teams that already use Datadog heavily now have a strong reason to consolidate their experimentation workflow into the same platform. Context-switching between a monitoring dashboard and a separate experimentation tool has real friction. A unified platform removes it.

For teams not on Datadog, or those with more complex data warehouse requirements, the competitive alternatives remain strong. But the market is clearly moving toward platforms that connect the code-to-business-impact loop end-to-end — and every major player in observability and analytics is now thinking about experimentation as a core capability, not an add-on.

Takeaway

Datadog acquired Eppo because it needed to close the gap between "monitoring what's broken" and "measuring what's working." Eppo's warehouse-native architecture and rigorous statistical engine were the right fit for a platform audience that already lives in data-rich environments.

One sentence summary: