Skip to main content

How to register custom conversion metric events in your experiment

Affected: All SDKs

Overview

If you are running experiments in LaunchDarkly, you can register custom conversion metric events to track contextual behavior and make data-driven decisions. By registering metric events, you can analyze the performance of different variations of your feature flag and optimize it accordingly.

Solution

To register events for metrics in experiments, you need two components: an Evaluation event, also known as a feature event, and a Metric event, also known as a custom event.  Refer to the documentation of Experimentation events for more information.

The Evaluation event is usually generated by a variation call associated with the feature flag used in the experiment. On the other hand, the Metric event is generally generated by a track call and connected to the experiment.
 
To emit a feature event, you can use either the variation() or variationDetails() methods. If you want to learn more about these methods, read Evaluating flags and Flag evaluation reasons.

Once the context has emitted a feature event, you can track it to your running experiment using the track() call. If you want to learn more about how to track, please read Sending custom events.

LaunchDarkly SDKs automatically flush pending analytics events to LaunchDarkly at regular intervals. However, you can manually call the flush() method to send events immediately without waiting for the next interval.  The flush() call is an optional step that may or may not be required, depending on your application. If you'd like to learn more about the flush call, read Flushing events.

Steps

  1. Evaluate a feature flag (for example: 'flag-experiment') with the variation() method

  2. Associate a running experiment (for example: 'A/B Testing') with the feature flag

  3. Connect a metric (for example: 'metric-ab-test') to the experiment with an event name (e.g., metric-event-name-ab-test)

  4. Track the metric using the event name.  For example track('metric-event-name-ab-test')

  5. [Optional] flush() call at the end

Resources

 

 

Was this helpful?