August 2020 Summaries

LaunchDarkly has become a Segment Destination, allowing businesses to make data-driven decisions using trusted source data. With this integration, users can start recording experiments immediately and measure results without any additional code or delay. Once the integration is installed, Segment events will automatically be associated with LaunchDarkly custom metrics. To get started, select "LaunchDarkly" as a destination in Segment's Destinations Catalog. For more information on how it works, visit https://docs.launchdarkly.com/home/experimentation/segment.

The Nano Series on "Empower" is the last in a series of four weeks discussing the Four Pillars of Feature Management, leading up to LaunchDarkly's Trajectory LIVE conference. Empower covers feature flag use cases that enable teams outside of Engineering to play a bigger role in delivering software. This includes using feature flags for feature entitlements, allowing customer-facing teams like Product Management and Sales to control which features customers have access to and when. It also involves sunsetting features by reducing technical debt and gracefully removing old features without needing much intervention from the engineering team. The series concluded with a discussion on how LaunchDarkly can be integrated into workflows, enabling different parts of a company to control feature behavior automatically or manually. Trajectory LIVE will offer two days of rich content on all things feature management and software development starting August 26th.

Empower is the fourth pillar of feature management that focuses on enabling customer-facing teams to control access to features, services, and products for customers. This approach reduces the burden on engineers and allows customer-facing teams such as Product Management, Marketing, Sales, and Customer Support to promptly deliver software to customers. Empower exemplifies progressive delegation, a core tenet of Progressive Delivery, which involves gradually transferring control of features to the team most responsible for their outcome. Software organizations can use feature management to empower all teams to influence software delivery in areas such as feature entitlements, customer targeting, proof of concepts (POCs) and trials, and sunsetting features. By leveraging feature flags, user targeting, and feature management platforms like LaunchDarkly, companies can improve operational efficiency, increase productivity across all teams, and enhance the overall customer experience.

The Learn pillar of feature management involves using feature flags to continuously learn from software and users. This includes conducting beta tests, running A/B/n experiments, setting baseline metrics, and measuring the impact of features on system performance. Dr. Claire Knight, Senior Engineering Manager at GitHub, discussed how teams can test their features with the right users, test for scale, create super minimal ships, and watch out for complexity while using feature flags. The Nano Series will continue to cover the Empower pillar next week.

An automated ticket distribution system in a support environment utilizes LaunchDarkly's percentage rollout feature, integrated with Zendesk, Heroku, and Redis to assign tickets to agents. Developed by Senior Support Engineer John Winstead, the system involves a Node.js script that processes ticket distribution via a "capacitor" model, which tracks distribution percentages for each agent. When tickets are generated, a webhook to a Heroku app adjusts these capacitor values based on pre-configured percentages, assigning tickets to the agent with the highest value, who then has their capacitor reset. This approach mitigates the randomness of traditional percentage rollouts, ensuring more consistent ticket distribution. The system, hosted on GitHub and connected to Zendesk's API, is poised for expansion to other internal automation tools, reflecting ongoing innovation in the team's support operations.

The Learn pillar of feature management focuses on enabling all teams to better understand how software changes affect users and systems. It is designed for the entire product delivery team, including developers, DevOps and site reliability engineers (SREs), and product managers. This pillar supports hypothesis-driven development and continuous improvement by integrating experimentation into the process of building software. Holistic experimentation delivers maximum value to customers while maintaining peak system performance and unifying engineers and non-engineers. It entails running not only A/B tests on public-facing features but also multivariate tests for back-end changes, collecting both quantitative and qualitative data, and aligning experiments with the deployment process. LaunchDarkly feature management enables holistic experimentation by allowing teams to manage releases and run experiments in one place, empowering product managers to release beta features, and seamlessly creating and running experiments including multivariate ones.

LaunchDarkly's Nano Series on the Four Pillars of Feature Management continues with an exploration of the Operate pillar, which focuses on improving operational health through feature flags. This includes using kill switches and circuit breakers, dynamic configurations, service metrics for reactive monitoring, and safe migrations. Michael McKay from IBM discussed these concepts in his talk "I Love Feature Flags." The Nano Series will cover the Learn pillar next week with a focus on shipping and learning fast via feature flags.

LaunchDarkly has introduced a new alternative form of patch called Semantic Patch to improve the handling of flag changes in their API. Unlike JSON Patch, which requires detailed knowledge of the current resource state and can be difficult to construct and debug, semantic patches allow users to describe higher-level changes they want to make to a flag. This improvement enables features like scheduled flag changes and relative changes to the flag's state. The existing flag PATCH endpoint will continue to accept JSON Patches, but sending a special header along with a flag PATCH request will interpret it as a semantic patch. Currently, anything that can be changed through the flag targeting page is supported by semantic patch, and more changes are planned to be supported soon.

LaunchDarkly is a feature management platform that provides Product Managers (PMs) with granular control over releases and reduced risk when introducing new features to customers. It allows PMs to precisely measure the impact of specific features, enabling them to release more value to customers more often. Key benefits include full control over feature rollouts, fine-grained targeting for who sees what experiences when, and the ability to measure the impact of your rollouts. LaunchDarkly also helps mitigate risks by providing kill switches and phased rollouts. Additionally, it extends its capabilities to support other teams such as sales, customer success, support, marketing, and design. Overall, LaunchDarkly enables organizations to deliver value to their customers more efficiently, confidently, and intelligently.

LaunchDarkly is a platform that enables teams to control their code's behavior and product features. It offers two types of flags: temporary and long-lived. Temporary flags are designed for short-term use, such as new feature rollouts or testing/hypothesis scenarios, while long-lived flags have an indefinite lifespan and support business logic or infrastructure operations. The platform allows users to create flags, define their behavior, and manage targeting rules based on user attributes and group memberships. It also provides tools for flag hierarchy management, code references identification, and automation of flag removal processes. By setting standards and following best practices, teams can effectively use feature flags to increase the safety and efficiency of software releases.

The "Operate" pillar of feature management focuses on improving the operational health of applications by making feature flags an essential part of operations. This approach reduces risk and ensures customer satisfaction, as seen in Honeycomb's low failure rate due to its use of LaunchDarkly. Key use cases for Operate include kill switches and circuit breakers, dynamic configurations, service metrics (reactive monitoring), and safe migrations. Feature flags enable teams to maintain system availability, adjust logging levels on the fly, and perform safer infrastructure migrations. By using a centralized feature management platform like LaunchDarkly, teams can better track key service metrics alongside feature flag events, improving overall stability, reliability, and availability of their services.