Compound Metrics - GrowthBook Documentation

Compound Metrics measure the share of exposed units who did every one of several things, in any order. For example, users who viewed pricing and started a trial. The result is a single conversion rate, analyzed like any other proportion metric. A compound metric is a Fact Table metric. Instead of a numerator, you define a list of components, each a Fact Table plus row filters.

Compound, funnel, or threshold?

Type	Use it for
Compound	Several distinct actions, order ignored, one number.
Funnel	Steps that must happen in order, with a rate per step.
Threshold	One action done at least N times, such as placed 3+ orders.

How a unit passes

A unit passes when it has at least one matching event for every component inside the metric window.

Order does not matter, and there is no co-occurrence window. Viewing pricing in week 1 and starting a trial in week 3 passes.
One event can satisfy several components if it matches all component’s row filters (e.g. event_name = view_pricing and plan = trial could be satisfied by a single event).
Units with no matching events fail.

Metric windows, metric delay, target MDE, and Bayesian priors work as on a proportion metric. A compound metric needs 2 to 20 components across at most 5 distinct Fact Tables, and no two components may be identical. In experiments, every component’s Fact Table must share an identifier type with the assignment table (or reach it through a Join Table); otherwise the metric is excluded from analysis. Not supported:

Use as activation metrics
Metric Slices are not supported.
Standalone metric analysis or Product Analytics

Compound metrics in experiment results

Results render like any other proportion metric. The metric drilldown adds a Components tab showing how many units in each variation matched each component. These counts are descriptive, shown for the overall result only, and carry no statistics.

Managing compound metrics with the API

Set metricType to compound and pass compoundSettings. Omit numerator and denominator; they are not allowed. The metric inherits its data source from the first component’s Fact Table.

curl -X POST 'https://api.growthbook.io/api/v1/fact-metrics' \
  -H 'Authorization: Bearer YOUR_API_KEY' \
  -H 'Content-Type: application/json' \
  -d '{
    "name": "Viewed pricing and started trial",
    "owner": "you@example.com",
    "metricType": "compound",
    "compoundSettings": {
      "components": [
        {
          "factTableId": "ft_events",
          "rowFilters": [
            { "column": "event_name", "operator": "=", "values": ["view_pricing"] }
          ]
        },
        {
          "factTableId": "ft_subscriptions",
          "rowFilters": [
            { "column": "plan", "operator": "=", "values": ["trial"] }
          ]
        }
      ]
    }
  }'

compoundSettings.components is a required array of 2 to 20 objects, each with factTableId and rowFilters. GET responses for compound metrics include compoundSettings and omit numerator. Listing fact metrics filtered by factTableId returns compound metrics with a component on that Fact Table.

​How a unit passes

​Compound metrics in experiment results

​Managing compound metrics with the API

How a unit passes

Compound metrics in experiment results

Managing compound metrics with the API