Modify experiment traffic - Amazon CloudWatch

Modify experiment traffic

Important

End of support notice: On October 16, 2025, AWS will discontinue support for CloudWatch Evidently. After October 16, 2025, you will no longer be able to access the Evidently console or Evidently resources.

You can modify the sampling rate for an experiment at any time, including while the experiment is ongoing. However, you can't update the treatment weights after an experiment is running. Therefore, you can change the total traffic exposed to the experiment after an experiment is running, but not the relative allocation to each treatment. If you modify the traffic of an ongoing experiment, we recommend that you only increase the traffic allocation, so that you don't introduce bias.

The following diagram shows how client traffic is allocated to different variations in an experiment. In this experiment, the sampling rate is 10% and the treatment weights for the two variations are 50% each.

A flowchart for the traffic allocation where the sampling rate is 10% and the treatment weights for the two variations are 50% each. 90% of the traffic goes to the default variation, and 5% goes to each variation in the experiment.
To modify the traffic allocation for an experiment

  1. Open the CloudWatch console at https://console.aws.amazon.com/cloudwatch/.

  2. In the navigation pane, choose Application monitoring, Evidently.

  3. Choose the name of the project that contains the launch.

  4. Choose the Experiments tab.

  5. Choose the name of the launch.

  6. Choose Modify experiment traffic.

  7. Enter a percentage or use the slider to specify how much of the available traffic to allocate to this experiment. The available traffic is the total audience minus the traffic that is allocated to a current launch, if there is one. The traffic that is not allocated to the launch or experiment is served the default variation.

  8. Choose Modify.