For the complete documentation index, see llms.txt. This page is also available as Markdown.

Variations

The Variations step is the core of your A/B test server-side campaign.

This is where you define what is actually being tested by configuring flags and their values.

Each variation represents a different experience exposed to users.

What is a variation?

In Feature Experimentation and Rollout (FE&R), a variation is defined by one or multiple flag values.

When a user is assigned to a variation, the SDK or Decision API returns the corresponding flag configuration.

This allows your backend to control exactly what experience is delivered.

Configuring a Variation

You can manage your variations directly from the interface:

  • Add a new variation

  • Duplicate an existing variation

  • Delete a variation

  • Add multiple flags within a variation

For each variation, you must define:

  • A flag key

  • A type

  • A value

Example:

  • Flag: button_color

  • Type: String

  • Value: orange

Configuration modes

You can configure your variations using two modes:

Standard mode

This is the default and most commonly used mode.

It allows you to:

  • Select a flag

  • Define its type

  • Assign a value

This mode is ideal for most use cases and provides a clear and structured interface.

Expert mode (JSON)

Expert mode allows you to define your variation using JSON.

This is useful when:

  • Working with complex configurations

  • Passing structured objects

  • Managing multiple parameters at once

Best practices

  • Keep variations simple and focused on a single hypothesis

  • Avoid testing too many changes in one variation

  • Use clear naming for each variation (e.g. “Control”, “New CTA”, “Variant B”)

  • Ensure flag values are correctly handled in your backend logic

Original vs Variation

Every campaign includes an Original variation. You have two options:

  • Override the original (recommended): Define a flag and value directly in FE&R. It avoids relying on hardcoded fallback values and keeps configuration centralized in the platform.

  • Use fallback values in your code: Do not configure the original in FE&R and rely on codebase defaults. It may result in less visibility and harder-to-maintain logic.

Next step

Once your variations are configured, proceed to Goals

Last updated

Was this helpful?