Skip to main content


Metrics serve as the foundation for evaluating experiments and determining the effectiveness of different variations. This overview provides a brief introduction to metrics in Eppo and highlights their significance in data-driven decision making.

Entities and aggregations

Metrics in Eppo are created at the entity level (such as a user), allowing you to define specific measurement criteria for different aspects of their experiments. An entity can be any meaningful unit of analysis, such as a user, session, or page view. To capture meaningful insights, metrics combine a fact and an aggregation. The fact represents the event or action being measured, while the aggregation defines how the data is summarized (e.g., sum, count, conversion, and retention) across the specified entity.

Ratio Metrics

Ratio metrics allow you to calculate ratios based on different metrics, providing deeper insights into the relative performance of variations. This enables a more nuanced analysis of experimental results, allowing businesses to understand the impact of changes in a more comprehensive manner.

For example, consider an average order value metric, which is created by dividing revenue (sum of prices) by number of orders (sum of items purchased or count of prices).

Average order value=RevenueNumber of orders\text{Average order value} = \frac{\text{Revenue}}{\text{Number of orders}}


Funnels are another powerful metric type in Eppo. They allow you to track and analyze multi-step processes or user flows. By defining the sequence of events in a funnel, you can identify bottlenecks, drop-off points, and conversion rates at each step. This facilitates a granular analysis of user behavior and helps optimize the customer journey.

Guardrail metrics

Furthermore, you can set metrics as guardrail metrics, which means they are automatically added to every experiment. This feature ensures that specific metrics are consistently tracked across experiments, providing a standardized measurement framework. By setting metrics as guard rails, you can maintain a unified approach to experimentation and easily compare results across different tests.