January 2024 Summaries

Decoupling deployment from release in software development can significantly improve speed and stability. This is achieved by using feature flags, which allow developers to control when a new feature becomes available to users after the code has been deployed. Feature flags enable gradual, targeted rollouts, laying the foundation for trunk-based development and continuous delivery. By separating deployment from release, organizations can increase their velocity, improve stability, reduce developer stress, and eliminate burdensome inter-team dependencies.

This tutorial demonstrates how to implement user targeting in an application built with Amazon Cognito and LaunchDarkly, using custom attributes from Cognito for creating segments and rules that impact the user experience. The example application showcases different experiences based on whether a user is identified as a "hobbyist" or "professional". It covers passing user data from Cognito to LaunchDarkly, setting up targeting rules in LaunchDarkly, and tying it all together for personalized digital experiences.

Autodesk, a global software leader, has increased its mobile app release velocity by 500% using LaunchDarkly. Previously, they released entire apps to all users at once, but now progressively deliver specific features to specific audiences. This approach minimizes customer-impacting software problems and ships valuable features to customers faster. By leveraging feature flags, Autodesk can decouple feature releases from app releases to the app store, allowing them to do more thorough testing and mitigate risk. The control LaunchDarkly provides has enabled Autodesk's mobile developers to release every 1-2 weeks, a 500% increase in release frequency.

LaunchDarkly's Migration Assistant helps organizations migrate cloud infrastructure, APIs, databases, and more by using feature management best practices. It allows for progressively rolling out back-end services to specific users, segments, and environments with context-aware targeting. The migration feature flags can be used in two-stage, four-stage, and six-stage migrations. These flags provide complete control over the migration process, making them more efficient and less risky.

LaunchDarkly is a platform that unites feature management and experimentation. It allows developers to create experiments on any software they're developing, from mobile apps to websites, across any part of the stack. The goal of these experiments is usually to determine what delivers the best results to end users. Experiments in LaunchDarkly involve defining a change, determining how its impact will be measured, and formulating a hypothesis. Once an experiment is set up, it can be executed by adding a feature flag and tracking user interactions. Metrics are then created to evaluate the success of the experiment. Finally, results from the experiment can be analyzed in the Results tab, which provides various segmentation options and probability charts.

This guide explains how to use feature flags effectively for targeting purposes. It introduces the concept of contexts, which can be anything and are used to determine targeting in LaunchDarkly. The user context is typically the most common type of context, containing details about the current end-user. Other types of contexts include organization, tenant, etc., allowing for more complex targeting rules. Multi-contexts enable creating multiple contexts simultaneously, and mobile contexts automatically add environment attributes to a mobile application and mobile device context. Secure mode allows setting all or specific attributes as private. Individual targeting sets a specific variation of a feature flag for a particular context instance by its key, while rule-based targeting targets any contexts that share an attribute or attributes. Segment targeting creates a segment of users and then targets the segment with variations. Targeting insights provide detailed information about contexts, segments, and flag variations received.