A/B Testing

A/B Testing with Traffic Distributor

A/B testing compares two application versions to determine which one produces the better user conversion rate. Traffic Distributor divides users between both versions and keeps each visitor on the same backend throughout the test.

Conversion can represent any important user action, such as purchasing, signing up, upgrading, subscribing, or clicking a required button. Visual design, marketing text, and other product characteristics can significantly affect the result.

On this page

Prepare Two Application Versions

1Create two application variants

Prepare two different versions of the same application or website for comparison.

Version AThe control version used as the baseline.
Version BThe alternative version containing the change to evaluate.

Configure Conversion Tracking

2Select a monitoring method

Use a suitable tracking method to record the desired user action for each application version.

  • A simple custom counter.
  • Application or server-side analytics.
  • A dedicated third-party testing utility.
  • A reporting tool with graphs and automatic conversion-rate calculation.
The monitoring method must distinguish the results generated by Version A and Version B.

Configure Traffic Distributor

3Install or reconfigure Traffic Distributor

Select the two application versions as its backends.

Traffic Distributor configuration for A/B testing
Configure Sticky Sessions and an equal 50/50 traffic ratio.
  • Routing method: Select Sticky Sessions.
  • Traffic ratio: Set 50 / 50 for a fair comparison.
Do not use Round Robin: A/B testing uses different backend content. Round Robin can move one user between versions and cause unavailable project elements.
For an already-running production project, integrate Traffic Distributor in front of it without interrupting normal application work.

Run the A/B Test

4Share the entry point

Provide users with the Traffic Distributor environment domain or a connected custom domain.

New visitors are divided evenly between both versions, while Sticky Sessions keeps each browser attached to its assigned backend.

Analyse the Results

5Compare conversion rates

Review the collected results and determine which version produces the higher conversion rate.

  • Compare total visitors for each version.
  • Compare completed target actions.
  • Calculate the conversion rate.
  • Confirm the result is meaningful before choosing a winner.

Use the Winning Version

Promote the better-performing application branch to production. The other version can be removed or modified for another test.

Traffic Distributor can remain in front of the production application to provide high availability and failover protection.

What’s next?