What is Eppo?
Eppo is a next-generation feature flag and experimentation platform built right on top of your data warehouse:
We offer companies an end-to-end suite of experimentation tools, but let you decide which ones are right for you. Here are a few common ways teams use Eppo:
- For experiment analysis alongside another feature flagging tool such as LaunchDarkly.
- For feature flagging without experimentation.
- As an end-to-end experimentation platform - both feature flagging and analysis.
How feature flagging works
Creating a feature flag in Eppo involves the following steps:
- Set up variations, with values that you control from the SDK
- Optionally, you can create fine-tuned targeting allocations that allow you to target subjects that match specific rules.
- Initialize the SDK.
- Embed the SDK in your code base.
- Enable the feature flag in your test or production environment.
- Log which experiments a user has been exposed to. You can pass in a logging callback to any SDK (e.g. a wrapper around Segment or Rudderstack) to route Eppo assignments to your data warehouse.
How experiment analysis works
Eppo's analysis is built on top of your data warehouse. Concretely, this means experiment results are computed within the warehouse without data ever leaving your system. As part of that process, intermediate and aggregate tables are always available in the warehouse for you to audit. In Eppo business metrics are defined in SQL, the same definitions that you use for business reporting.
Generating an experiment report on Eppo involves five pieces:
- Use your feature flagging tool of choice to send experiment assignments into your data warehouse.
- Connect Eppo to your data warehouse. Currently Eppo supports Snowflake, Redshift, BigQuery, and Databricks.
- Annotate experiment assignments and event streams by writing short SQL snippets.
- Monitor your experiments' progress.
- Explore and share experiment results.