August 2021 Summaries

Becoming a frontend engineer can follow multiple paths, including obtaining a degree, self-teaching, or attending bootcamps, as shared by Christina Chen. She chose frontend engineering due to her preference for client-side coding and the opportunity to work closely with UX designers and end-users. In her role, she collaborates with a team that includes engineers, a product manager, and designers, focusing on writing clean, maintainable code and automated tests. Contrary to popular belief, coding occupies a small portion of her time; much is spent on brainstorming and testing. Weekly routines involve standups, team meetings, and planning sessions, while occasional events include retrospectives, hackathons, focus weeks, and social activities. Her experiences highlight the variety and collaboration inherent in a frontend engineering career.

Product managers play a crucial role in creating clear differentiation and competitive advantage for their products. They are responsible for various aspects, including pricing strategy, promotion, and placement. Working cross-functionally with engineering, design, marketing, and revenue teams, product managers face immense pressure to succeed. However, stress can lead to inaction and hesitation to take risks. Experimentation is a proven method for developing successful products. Companies like Booking.com run thousands of concurrent tests to iterate their way towards success. To foster an environment of healthy risk-taking, four fundamental practices can be implemented: 1. Chaos Engineering: Introduce failure as a conscious constant by using tools like the "Chaos Monkey" to stress test systems and improve incident response skills. Discuss outcomes in direct and explicit language, acknowledging experiments as learning experiences. 2. Blue-Green Deployment: Create two nearly identical instances of an application behind a load balancer for continuous delivery with minimal downtime. This method allows testing and experimentation in a production environment without sacrificing uptime. 3. Canary Deployment: Test changes and trial releases on a small group of users to gather real-world data and adjust deployment strategies accordingly. 4. Review Recovery: Accept the inevitability of mistakes and create a safe, judgment-free space for people to share their missteps. Encourage learning from others' experiences and foster open communication about failures. By incorporating these practices, organizations can cultivate a culture of experimentation that uses failure as a catalyst for future successes.

LaunchDarkly recently hosted a brunch-and-learn event with customers such as Tabcorp and Bank of New Zealand to discuss how Asia-Pacific companies are using feature management and LaunchDarkly to drive business. The discussion focused on the primary use cases for feature management, highlighting the benefits both companies have experienced using LaunchDarkly. Both organizations operate in highly regulated industries where speed and flexibility are crucial while maintaining control under stringent local and federal oversight. Using LaunchDarkly's feature flagging platform has allowed them to deliver new functionality to customers without stress or regulatory issues. Additionally, the conversation touched on the challenges of building in-house feature management solutions and the benefits of using a well-architected tool like LaunchDarkly instead.

Multi-armed bandit algorithms are a class of algorithms designed to minimize regret when running experiments, such as A/B tests. These algorithms allocate more traffic to the best performing variation as they learn more about the value of each variation. This approach is particularly useful for marketing site changes where maximizing conversions is important. LaunchDarkly supports both automated and non-automated multi-armed bandit implementations, with the latter involving manual adjustment of traffic allocation based on experiment results. The epsilon-greedy algorithm is an example of a multi-armed bandit algorithm that balances exploration and exploitation of variations.

LaunchDarkly introduces a new feature that enables users to sync audience data from digital optimization tools like Amplitude into a LaunchDarkly segment, allowing for seamless integration and efficient management of user groups. This eliminates the need for manual importing or updating of target audiences on a schedule for feature flagging. The process involves creating a cohort within Amplitude, syncing it to LaunchDarkly, and then using the synced segment in feature releases. Enterprise plan customers can access this feature now, with plans to support additional data tools and create Big Segments directly within LaunchDarkly in the future.

LaunchDarkly, a feature management platform founded by John Kodumal and Edith Harbaugh seven years ago, has raised $200 million in additional funding. The company is featured on the 2021 Forbes Cloud 100 list. Serving over 2,000 customers worldwide, LaunchDarkly's platform supports more than 20 programming languages and processes peaks of over 20 trillion feature flags daily. The platform provides technical value by supporting various industries, business value through continuous delivery without significant downtime, and personal value by allowing safe feature releases and management. With a growing team of over 300 employees worldwide, LaunchDarkly continues to invest in its scalable feature management platform and is currently hiring for various roles.

LaunchDarkly has introduced a major update to its experimentation add-on, making it easier to build and manage experiment audiences. The new experiment allocation rule helps prevent biased audience selection and allocation, prevents carry-over bias when ramping down variations, enforces consistent user experience for experiment audiences, provides easy control over experiment event volume costs, and assists team members in completing experiment setup. To access these features, users must update their SDKs to the latest versions and contact [email protected].

LaunchDarkly, a feature management platform, has been recognized by Forrester as a leader in their New Wave report on Feature Management and Experimentation. The company's approach to Progressive Delivery includes managing features, enabling safe releases, and automating rollouts. Forrester highlights the importance of feature management and experimentation tools for DevOps teams aiming to create faster release cycles while staying connected with customers. LaunchDarkly plans to continue innovating in this space by improving user opt-in and accessibility options.