December 2024 Summaries

We’re thrilled to share that LaunchDarkly has been recognized as a leader in the G2 Winter 2025 Reports for Feature Management, earning several #1 rankings and badges across key categories. This recognition highlights how the LaunchDarkly platform can help teams release confidently, experiment effectively, and innovate faster by providing measurable results, reducing risk, and improving user experiences. Earning top spots on the G2 Grid validates LaunchDarkly's ability to deliver transformative software delivery processes for its customers. Customers praise LaunchDarkly for its ease of use, powerful experimentation tools, seamless feature flag management, and real-time analytics. By giving teams the tools to ship faster with confidence, learn through experimentation, and deliver at scale, LaunchDarkly helps them succeed in their software development endeavors. The company is honored to receive this recognition from G2 and deeply grateful to its customers for sharing their experiences.

To add LaunchDarkly feature flags to a Ghost CMS, one must first create a hosted Ghost account and obtain a free LaunchDarkly account. The user then needs to paste a specific code snippet into their site's footer tab in the Ghost admin panel, which uses LaunchDarkly's JavaScript SDK to check if a dark mode flag is enabled. When the flag is on, it changes the site's styling by modifying various CSS properties. A new feature flag is created in LaunchDarkly with a custom configuration, and its client-side ID is copied and pasted into the code snippet. The user can then toggle the flag on or off to test different variations of their website's design. This technique allows for easy experimentation and personalization of content without requiring extensive coding knowledge.

At LaunchDarkly, the introduction of Frequentist and Bayesian statistical models gives teams the flexibility to choose their preferred approach when designing experiments. This allows teams to work more collaboratively by focusing on interpreting results and driving action, rather than debating methodologies. The tool now provides a toggle for selecting between Frequentist and Bayesian analysis types, along with enhancements such as improved results presentation and a sample size calculator. With this flexibility, LaunchDarkly aims to make experimentation more accessible and flexible, allowing teams to trust the results of their experiments and eliminate friction in experimentation.

In LaunchDarkly, flag targeting rules allow for hyper-specific targeting of specific individuals or groups, enabling hyper-targeted experimentation. Flag targeting rules evaluate users based on predefined criteria and serve variations accordingly. The order of flag rules matters, as the top rule evaluates first, followed by subsequent rules in sequence. This allows for complex targeting scenarios to be achieved through a combination of separate and combined rules. Targeting rules can also be used to capture margins by leveraging user attributes attached to the context and experiment learnings. By adding new targeting rules after running an initial experiment, users can quickly capitalize on winning margins and improve their experimentation program's efficiency. Effective use of flag targeting rules requires careful consideration of rule clarity and targeting criteria to ensure optimal results.

In an effort to stay competitive in the market, Ally Financial, the largest all-digital FDIC-insured bank in the United States, embarked on a modernization plan to improve its software release process. The company's Director of Digital Engineering Operations and Director Lead of Software Engineering worked with LaunchDarkly to build faster and safer releases. Ally's transformation began about 15 years ago when it transitioned from traditional banking services to digital, resulting in the organization it is today. With over 11 million customers and nearly $200 billion in assets, Ally is a powerhouse in fintech. To balance speed and safety, Ally modernized its applications, introduced feature-level rollbacks, implemented testing in production, and achieved parity between web and mobile platforms. The company collaborated with LaunchDarkly to implement feature flagging capabilities, decoupling the UI from the API and introducing a more modular architecture. With these changes, Ally reduced overnight and weekend releases by 97% while increasing production deployments by 300%. The team developed guiding principles for release management, emphasizing autonomy, automation, and continuous improvement. Ally's success with LaunchDarkly has sparked interest across the organization, and the company plans to continue improving its software delivery process and work-life balance of engineering teams.

Software release management is a complex process that involves both technical and cultural aspects. Despite advances in technology, challenges persist for software delivery teams. Companies should prioritize user experience, learn from failures, implement strategies like canary releases and feature flags to minimize fallout, and ensure the ability to ship software on any day without risk. Metrics and observability are key for monitoring performance, while AI introduces new risks that require robust testing and observability. Lastly, fostering a culture of excellence is crucial for successful software delivery management.

Software deployment involves delivering new products, services, and features to users safely, efficiently, and continuously. A well-architected toolchain automates releases, minimizes deployment risks, and helps teams recover quickly when issues arise. Key components of a modern software deployment stack include version control systems like GitHub, GitLab, and Bitbucket; cloud deployment platforms such as AWS CodeDeploy, Azure DevOps, Google Cloud Deploy, and Heroku; CI/CD tools including Jenkins, CircleCI, GitHub Actions, and GitLab CI; configuration management solutions like Ansible, Puppet, and Chef; and feature management platforms like LaunchDarkly, Split, Flagsmith, Unleash, ConfigCat, and Firebase Remote Config. When choosing the right software deployment tools, consider scale and performance needs, team capabilities and resources, and essential integrations. Building an effective software deployment pipeline involves starting with a solid foundation, automating strategically, implementing progressive delivery, monitoring and responding to issues, and refining and iterating over time. Nine best practices for modern software deployment include starting small and iterating, automating everything you repeat, making rollbacks easy, monitoring deployment health, keeping deployments small, documenting deployment requirements, testing in production-like environments, planning for failure, and practicing progressive delivery. Feature management platforms like LaunchDarkly help teams innovate with confidence by separating code deployment from feature release, enabling control over software delivery, gradual feature rollouts, real-world feedback gathering, and quick issue responses without redeploying.

Downtime in software releases can cost organizations millions, making a comprehensive release management checklist essential. Such a checklist should include pre-release planning, development and testing, release preparation, execution, and post-release phases. It should also incorporate modern feature management platforms like LaunchDarkly to add layers of control and safety. A well-structured release management checklist can minimize downtime risks, streamline the release process, and improve overall application stability.

LaunchDarkly has introduced enhancements to its Snowflake data export, allowing customers to perform advanced custom analysis within their existing data tools and workflows. This update enables engineering, product, and data teams to create experiments in LaunchDarkly and export the audience data for further analysis directly within their Snowflake workflows. The enhanced data export provides more flexibility and control, enabling precise audience assignment and leveraging preferred analysis tools. It also allows custom warehouse analysis, empowering data-informed product decisions that drive greater business impact.