Create a new experiment

This article helps you:

Create and initialize a new experiment
Add context to your experiment description, so other stakeholders will understand it

The decisions you make in the design phase set the stage for your experiment’s success. By putting more thought into your experiment’s purpose and goals before you start, you’ll be far more likely to glean useful, actionable insights from it.

To create a new experiment, install an SDK or call the evaluation REST API. Then follow these steps:

Click Create > Experiment, and select an Web or Feature.
In the New Experiment modal, complete the fields:
- Name: Enter the name of the experiment for future reference.
- Project: Select the project in which this experiment operates.
- Experiment Type: Select from the following:
  - Hypothesis Testing (default): Experiments where you’re using data to decide which variant to roll out based on performance. If no variant outperforms the control, you’ll usually want to roll back the experiment and stick with the control experience.
  - Do No Harm (DNH): Experiments where you already have a direction in mind, and the purpose of the experiment is to make sure that this change doesn't significantly harm key metrics. This type of experiment is often used for design system changes, or features that have to be sunset.
  - Multi-Armed Bandit: Amplitude allocates an increasing amount of traffic to the winning variant, based on the primary metric, until it hits 100% allocation.
- For Web Experiments, enter the Targeted Page URL, on which this experiment runs.
Optionally, complete the following fields:
- Key: Keys are unique to experiments and tell which experiments a user participates in. You can edit keys until you run the experiment.
- Evaluation Mode: Select if the experiment runs locally or on Amplitude's Experiment servers. For more information, see Local evaluation and Remote evaluation
- Bucketing Unit: Select the unit Amplitude uses to assign variants, either User or Group.
Click Create.

For example, you've chosen to run a hypothesis testing experiment with a direction setting of "increase" and a minimum goal (MDE) of 2%. This means you believe the metric should increase by at least 2%. If you change the experiment type to Do No Harm, you expect the metric to "not increase by 2%." A good use case for a Do No Harm experiment is launching a service agreement in your app and then testing for a lack of change in user retention.

Click Continue to move on to the next step—defining your experiment’s goals.