We are excited to announce the public preview of Split Experimentation for Azure App Configuration.  In today’s software development, delivering high-quality features rapidly while minimizing risk is a top priority. Our new experimentation capability extends feature flags in App Configuration, helping you balance speed, accuracy, and safety to effectively de-risk application development.

This new capability leverages the existing services you utilize to host your applications in Azure, integrating seamlessly with App Configuration for feature definitions and Application Insights for telemetry data. The experimentation engine is powered by the Split Experimentation workspace, a new Azure native ISV service that incorporates robust analysis capabilities from Split Software, Inc.

Why Experimentation?

Feature flags in App Configuration already offer substantial benefits, including risk reduction, enhanced operational control, and targeted user segmentation. Experimentation takes these benefits to the next level, unlocking major value in the form of:

Data-Driven Decisions: Optimize features based on actual user data, improving performance and satisfaction.
Risk Mitigation: Identify potential issues early by exposing features to a controlled group, reducing widespread problems and improving MTTD and MTTR.
A/B Testing: Determine which feature version performs better through controlled user group testing.
Faster Feedback Loop: Quickly iterate and improve features based on real user input.

In this era of intelligent applications and AI-driven technologies, experimentation plays a pivotal role in harnessing the full potential of AI models. It enables developers to effectively integrate AI models, optimize user experiences, and significantly enhance the success of Gen AI based features in their intelligent applications.

How to run experimentation in Azure?

Azure App Configuration now offers Variant Feature Flags, enabling more granular control and flexibility beyond traditional Boolean flags. Here’s how to run experiments on these variants in three simple steps:

Define your Feature: Specify your feature and its variants in Azure App Configuration to provide tailored experiences for different scenarios.
Send Telemetry Data: Send telemetry data on variant evaluations and events to Application Insights to monitor performance and impact.
Experiment and Define Metrics: Create experiments on variant feature flags and define metrics to track the new features.

Split Experimentation leverages telemetry data from Application Insights to pinpoint the top-performing variant based on your defined metrics. Experiment results can be viewed directly in the App Configuration store.

Getting Started

Explore our sample application to see how experimentation can enhance your development process. Start with the .NET azd sample, Quote of the Day, featuring a variant feature flag “Greeting” with two variants: “Off” and “On.” Users with the “On” variant see a greeting before the quote; users with the “Off” variant see only the quote.

Off & On variants in UI shown side by side

Success is measured by tracking the “Like” metric for the click event on the heart button, showing user preference. Results show that “On” variant is the desired variant, that is, users tend to click on heart button more if they see the greeting with the quote.

Experiment results showing outcome and impact for defined metrics

Learn more

For more details, check the documentation. Join the community of developers transforming their feature management strategies with Azure and Split.io. Happy experimenting!

​Microsoft Tech Community – Latest Blogs –Read More