Restart an experiment

Restarting experiments can be useful if you had instrumentation issues that resulted in data quality issues and have fixed them. When you restart an experiment, you exclude any previous user data from the monitoring and analysis of your restarted experiment.


This feature is in Beta. If you would like to try out an early version, contact Amplitude Support.


There are two primary requirements to be able to restart your experiment.

  1. Your experiment must use Amplitude's default exposure tracking.
  2. Your client SDK version must support experiment restarts.
SDK Minimum version
JavaScript v1.8.0
Android v1.10.0
iOS v1.11.0
React Native v1.2.0

Options when restarting

The following changes will be made to your experiment configuration when restarting:

Property How it will be changed after restart
Experiment Key Updated to a new value
Exposure Event If your experiment is using a custom exposure event, it will be reset to the default Amplitude exposure event
Start Date Updated to the date of the restart
End Date (Optional) Updated to your selected value
Bucketing Salt (Optional) If selected, randomized to a new value
Sticky Bucketing If you selected the option to re-randomize users and sticky bucketing is on, it will be turned off
Decision If you previously rolled out or rolled back your experiment, this decision will be erased

Experiment Key

The experiment key property is used by Amplitude exposure events to differentiate when your experiment has been restarted.

When using the Evaluation API , the response body will contain the experiment key of the currently running experiment.

2 "<flag_key>": {
3 "key": "<variant_value>",
4 "payload": <variant_payload>,
5 "experiment_key": "exp-1",
6 },
7 // ...

Similarly this experiment key is available when using fetching variants using Experiment SDKs.

For example, for the JavaScript SDK:

The Variant object will contain the existing value and payload properties, along with a new expKey property.

After restarting

After the experiment restart, the exposure event will be updated to use the new experiment key.

A sample exposure event

This new filter on the exposure event ensures that previously recorded exposure and metric data will not be included in your experiment analysis.

Was this page helpful?

Thanks for your feedback!

May 31st, 2024

Need help? Contact Support


Have a look at the Amplitude Blog

Learn more at Amplitude Academy

© 2024 Amplitude, Inc. All rights reserved. Amplitude is a registered trademark of Amplitude, Inc.