May 2025 Summaries

The Experiment Decision Framework (EDF) is designed to aid experimenters in making informed decisions about their experiments by automating the determination of when sufficient data has been collected and recommending actions based on predefined business criteria. By setting Target Minimum Detectable Effects (MDEs) and decision criteria upfront, the framework addresses issues such as false positives and inconsistent decision-making due to business pressure or personal bias. It provides automated status updates indicating whether an experiment should continue, be altered, or be concluded based on statistical precision and predefined goals. The EDF is available for Pro and Enterprise customers and integrates with existing experiment data and statistical engines, allowing for customizable settings per experiment while maintaining a structured approach to decision-making without altering the experiment's operational process.

It’s estimated that over 40% of all new code is now written by AI and this trend is expected to accelerate, leading to a significant increase in code being shipped to production. However, the impact of this increased code on companies' core business metrics is uncertain and may not always be positive. Industry-wide, only about a third of product changes have a beneficial impact, while the rest either do nothing or can even be harmful. To mitigate this issue, feature flags and experiments have been used by top companies for years to measure code changes in production and ship successful ones while reverting unsuccessful ones. In response to the increasing scale of the problem, GrowthBook has launched two products tailored specifically for this new reality: Safe Rollouts, which enable lightweight experiments with minimal developer involvement, and the MCP Server, a feature management and experimentation platform that allows developers to create feature flags, configure Safe Rollouts, and monitor results all within their IDE, making it easy to incorporate GrowthBook into AI-driven development workflows.

You're searching for a flag related to the new settings menu, but can't find it due to hundreds of flags created by different team members. However, with Search Filters, you can narrow down hundreds of flags to just 15 that match your criteria, such as type, rule, status, and more. This feature is not only useful for one-off searches but also helps wrangle flags across teams, making it a faster and cleaner way to manage them. Additionally, power users can utilize the real-time syntax updates in the search box, allowing them to skip the UI and explore further query formats and examples. The Search Filters feature is now live on Cloud and will be included in the next release.

The GrowthBook MCP Server is a tool that enables developers to integrate AI coding tools with the GrowthBook platform, allowing for easy addition of feature flags, safe rollouts, and A/B tests without leaving the code editor. This integration is made possible by the Model Context Protocol (MCP), which formalizes "tool calling" between AI tools and platforms. The MCP Server supports 14 tools, including popular ones like VS Code, Cursor, and Claude, and can be easily set up with a JSON configuration file. As support for features like Resources and Prompting grows, the GrowthBook MCP Server will continue to evolve and improve, and is already available for use in various code editors and platforms.

The latest release of GrowthBook Cloud introduces several exciting features, including Safe Rollouts, a new type of feature flag rule that allows for gradual release of features while monitoring guardrail metrics for regressions. This release also includes a Time Series view for metrics in experiment results, enabling users to track changes over time without requiring additional SQL queries. Additionally, an official Jira integration has been launched, allowing users to link features or experiments to Jira issues and access key details directly within Jira. The release also extends the Experiment Decision Framework to the REST API and Webhooks, providing recommendations for experiment decisions and enabling custom workflows. Furthermore, the GrowthBook Dev Tools browser extension now supports Server-Side Rendered JavaScript applications, making it easier to debug and QA feature flags on the back-end. Overall, this release aims to enhance the user experience and provide more functionality for developers and product managers, with several more features planned for the near future.