Stopping an experiment
You can stop an experiment run at any time. Common reasons for stopping include:
Operational issues, invalid experiment behavior, or unexpected treatment performance: Stopping the experiment run prevents additional users from being exposed to treatments and returns users to the currently deployed feature flag configuration.
A clear treatment winner: Results consistently meet the success criteria and operational metrics remain within acceptable thresholds. Before promoting a winner, verify that the treatment performs better than the control for the intended outcome. For information about rolling out the winning treatment, see Promoting a winning treatment.
Inconclusive results or changing priorities: The experiment is no longer relevant, the hypothesis needs revising, or you want to restart with a different configuration.
Use the following procedure to stop an experiment.
To stop an experiment
Open the Amazon Systems Manager console at https://console.amazonaws.cn/systems-manager/appconfig/
. -
In the navigation pane, choose Experiments, and then choose a running experiment. The experiment dashboard opens.
-
Choose Stop.
-
(Optional) In the Experiment results summary section, document the experiment outcome to help your team make an informed launch decision. This summary is saved with the experiment run for future reference. You can provide:
Executive summary – Summarize the experiment outcome and key findings.
Reasons to launch – Evidence in favor of launching the treatment.
Reasons not to launch – Evidence against launching the treatment.
-
In the Confirmation section, type
confirm, and then choose Stop experiment run.