Configure your experiment

Configuring Feature Experiment is a two-stage process:

• Creating a deployment
• Installing the SDK you want to use.

Configuring Experiment isn't the same as creating an experiment. You must first configure Experiment and ensure that it's working before you can create any experiments. This page is for Feature Experiment. If you want to configure Web Experiment, go to Implement Web Experiment.

Creating a deployment

In Experiment, a deployment is where you can serve a group of flags or experiments for code execution. After you create a deployment, Experiment generates an access key, which you can then use to properly route your flags and experiments.

Deployments live under Amplitude Analytics projects. A project can have multiple deployments, but you can attach each deployment to a single project.

To create a deployment follow these steps:

1. Go to Experiment > Deployments and then click Create Deployment.
2. Choose the Amplitude Analytics project to which you want to associate the deployment. If you want to create deployments in multiple projects, select all the relevant projects from the drop-down list.
3. Enter a name for your deployment. Names should be descriptive of the work you are doing.
4. Specify its type:

• Client-side: These deployments run on a client device, such as a web browser or mobile app. The deployment key associated with client deployments is publicly viewable and should be used in client-side SDKs.
• Server-side: These deployments run on a server you control, such as a web server or batch processing system. Keep the deployment key associated with server deployments secret and use it server-side SDKs. Server-side keys can access the REST API for flag evaluation. If, instead of running a full-blown experiment, you only need to evaluate flags using the REST API, you should create a server-side deployment.

5. Click Create a Deployment. Experiment creates your deployments and automatically generate keys to copy and use.

Deployment keys

Both types of deployments generate keys that are specific to the deployment. A key is a unique identifier for the deployment that ensures data integrity. The key ensures that all data collected by your experiment is associated with that particular deployment and all results and analytics based off of those results are as accurate as possible. For more information about keys, go to Keys and Tokens.

• Client-side keys: The deployment key associated with client deployments is publicly viewable.
• Server-side keys: Keep the deployment key associated with server deployments secret and use it only for server-side SDKs. Server-side keys access the API for flag evaluation.

Install the SDK

If you don't use the API, the next step is to install an Experiment SDK. All SDKs send a request to Amplitude Experiment to decide what flag configurations it should serve to a particular user. There are some important differences between client-side and server-side SDKs you should be aware of.

Client-side SDKs run in the end-user application deployment. For client-side SDKs:

• Assume a single user deployment.
• Use client-side deployment keys, which are public and visible to end users.
• Fetch variants up front for a given user.
• Store variants locally on the client for offline mode.

Server-side SDKs run in a server deployment. Server-side SDKs:

• Assume a multi user deployment.
• Use Server-side deployment keys, which you should keep private.
• Fetch variants on each request.

The User context

When assigning variants, the evaluation engine applies the targeting rules to a user context object, which represents the identity of an individual user. In client-side SDKs, this object-user relationship is set on initialization and passed to the server on every request for variants. In server-side SDKs, the user may change, and is set on every request.

When targeting individual users to assign variants, Experiment matches on any of the listed user identifiers, such as user_id and device_id. Using rule-based user segments, users match on any of the predefined properties (country, platform, and so forth), or on custom properties specified in the user_properties object. Read more about defining experiment users in this article.