Ship Fast, Break Nothing: LaunchDarkly’s Winning Formula
In today’s rapidly evolving software landscape, the ability to deliver code quickly while maintaining stability has become a critical competitive advantage. LaunchDarkly, founded in 2014, has emerged as the leader in the feature management space, transforming how companies deploy software by separating deployment from release.
The company celebrated its 10th anniversary last July and will hold its next annual Galaxy user conference in mid-May.
What’s in a Name?
The name LaunchDarkly comes from the software development technique of launching darkly (also called “dark launching”), which is the practice of deploying new code or features to production environments while keeping them invisible to most or all users. This approach allows developers to deploy code without immediately exposing it to users; test features in production with limited audiences (like internal team members) before wider release; gradually roll out features to increasing percentages of users to monitor performance and impact; and quickly disable problematic features without rolling back entire deployments.
What Is Feature Management?
At its core, feature management allows development teams to implement feature flags: conditional statements that control how code behaves in production without requiring new deployments. Feature flags are “features turned on/off during runtime without deploying new code” that provide better control and enable experimentation.
In its simplest form, feature management allows companies to separate deployment from release. Before feature management, when companies deployed software changes, their entire customer base would immediately see the latest version. If something went wrong, the impact could be widespread and damaging.
With feature flags, companies can deploy software but gradually increase its release, starting with just 1% of users, then 2%, then 5%. If problems arise, the “blast radius” of errors is significantly reduced, affecting only a small portion of users instead of the entire customer base, said Claire Vo, chief product officer at LaunchDarkly.
The Birth of a Category
According to LaunchDarkly executive chair and Co-founder Edith Harbaugh, before founding the company, she observed that many organizations were cobbling together homegrown systems for feature flagging. Netflix had one, Amazon had one, but there was no commercial solution available on the market. LaunchDarkly became the first platform that made feature management accessible to any company.
Meanwhile, CTO and co-founder John Kodumal – who was a classmate of Harbaugh at the prestigious Harvey Mudd College — came from Atlassian, where he had experienced the limitations of an internal feature flagging platform. The tool was painful to use, took almost a year to develop, and had minimal ongoing investment, Kodumal told The New Stack. This gap in the market became the inspiration for LaunchDarkly, he noted.
Moreover, the company’s name itself was carefully chosen. As Harbaugh explains, “People were using different words for it. Dark launches, dark rollout. I liked LaunchDarkly because it was a little bit more positive.”
LaunchDarkly co-founders Edith Harbaugh and John Kodumal.
Beyond Simple Feature Toggles
LaunchDarkly has evolved well beyond basic feature flags to become a comprehensive release management platform. Current CEO Dan Rogers — Harbaugh, though an engineer by training, was the company’s first CEO — describes the company’s offering as encompassing four key areas:
- Feature management: Controlling the software experience users receive
- Product experimentation: Bringing A/B testing capabilities to engineers and product teams
- Release monitoring: Measuring software performance at the point of release to detect issues before they affect all users
- AI engineering: Applying feature management principles to AI features, allowing companies to control which models and prompts are used and to measure their effectiveness
Feature Management in the Federal Government
For federal agencies undertaking modernization efforts, feature management has emerged as a critical capability. This approach helps government organizations reduce risk in software releases while enabling innovation.
Sara Mazer, LaunchDarkly’s field CTO and federal sector representative, told The New Stack: “Feature management is really separating out deployments from feature releases. And so, engineers can keep working and submitting their code and checking things in as it completes. And then at some point in the future, they or a nontechnical program manager can enable that feature.”
This separation is especially valuable for government agencies transitioning from monolithic systems to a microservices architecture. Feature management provides the granular control needed to support the principle of least privilege, allowing for safe, gradual rollouts of new features, Mazer said.
The risk reduction aspect is particularly appealing to federal agencies, she said. LaunchDarkly’s representative notes that feature management “reduces risk for their agency. It even fits into the security, zero trust architecture, and always expecting that there’s a breach. So, you can use that kill switch to turn something off that is related to your breach without killing your whole entire application,” Mazer added.
This capability allows agencies to minimize impact on users while quickly terminating potential security threats, aligning with modern security approaches like zero trust architecture.
LaunchDarkly made an early investment in obtaining FedRAMP authorization, recognizing that the public sector has a critical need to control release risks where the consequences of buggy features could significantly impact citizens or government operations.
Industry Analyst Perspective
Industry analysts view feature flags as increasingly critical for modern software development. According to IDC analyst Jim Mercer, feature flags are effective in helping development teams release software faster by decoupling feature deployment from feature release, allowing teams to deploy code more frequently without impacting users. They enable gradual rollouts, where teams can monitor performance and identify issues before a full-scale launch, reducing the risk of production outages.
Mercer also notes that feature flags play a key role in product marketing through experimentation via A/B testing, enabling data-driven decisions about product features. He considers them critical for modern digital businesses seeking to improve agility while mitigating the risks of rapid changes.
Thomas Murphy, an analyst at Gartner, highlights that feature toggles are particularly valuable when you have the right underlying application architecture. They’re a great way to “derisk a deployment” through progressive releases, let businesses test ideas, provide ways to “test in production,” and can be more accurate for performance testing.
Moreover, feature flags are useful “if an organization is making use of lots of microservices, calling external APIs, and has a high release cadence, the ability to test an application and know that it is still good 15 minutes later,” Murphy said. Yet, “they are also not the only thing you will want — you need APM/observability tools and a plan for where and how you use them, you have to manage flags and make sure old ones are retired, etc.,” he added.
And in the wake of major tech outages, RobustCloud analyst Larry Carvalho observed that by integrating feature flags into development processes, teams can innovate faster with lower risk.
“With the increased attention to robust testing after a blackout, this feature may gain traction among the developer community,” he told The New Stack.
Enterprise Adoption and Growth
The company has experienced significant growth, now serving more than 5,500 enterprise customers, including almost a quarter of the Fortune 100. Their platform processes a staggering volume of feature flag evaluations — 41 trillion on a peak day last year —demonstrating both the scale of adoption and the robustness of their infrastructure.
Companies across various industries, not just high-tech firms, are adopting LaunchDarkly’s platform. As Rogers told The New Stack: “When companies realize that software is powering the world, and therefore, the work that software developers are doing is critical to them, that’s really where we have a great opportunity.”
Competitive Landscape
The feature management market has evolved significantly since LaunchDarkly pioneered it. IDC’s Mercer notes that most independent feature flag companies have been acquired. “Rollout was acquired by CloudBees in 2019, and Split was acquired by Harness in May 2024. Only a few small players remain, such as DevCycle, Optimizely, and Taplytics,” he told The New Stack.
Despite increasing competition, including open source options, LaunchDarkly has maintained its leading position. Mercer attributes this to the fact that LaunchDarkly was the first to commercialize feature flags and built a strong brand name in the space, focusing on creating a feature management platform with diverse capabilities that cater to various customer needs, including scaling features across enterprises.
“First mover advantage gave them a good lead, they did a good job of raising capital,” Gartner’s Murphy told The New Stack.
However, Murphy added that the feature flag market has become less unique over the last decade, with many options now built into DevOps pipelines.
“I think the development of OSS [open source software] solutions and the integration of flags into DevOps pipelines is making it slowly less of a standalone market, but for very complex scenarios, LaunchDarkly is a complete solution,” he told The New Stack.
Recent Innovations
LaunchDarkly continues to expand its platform capabilities. At their Galaxy ’24 user conference in May 2024, the company unveiled several new features focused on helping teams “ship at the speed of now,” including:
- Enterprise release management and automation through Release Assistant
- Release Guardian, which helps teams identify and fix software performance issues before they become customer-facing problems
- Advanced Experimentation capabilities
- LaunchDarkly for AI Engineers, including AI Prompt and Model Templates
Last November, LaunchDarkly announced a new approach called “Guarded Releases” that aims to change how organizations deliver software. This strategy incorporates application performance thresholds, release auto-remediation, and release monitoring to help teams ship quickly without compromising stability.
LaunchDarkly also introduced “Guardian Edition,” its offering that brings together the core elements of Guarded Releases along with Error Monitoring and Session Replay capabilities.
Guardian Edition helps fill the gap for organizations struggling with increasingly complex software builds by providing “integrated real-time insights to close the feedback loop, automating the end-to-end process of Progressive Delivery,” said James Governor, principal analyst and co-founder of Redmonk, in a statement.
The company also introduced AI Configs, enabling engineering teams to manage large language models (LLM) and prompts at runtime, allowing teams building AI software to quickly iterate on prompt and model configurations while tracking performance metrics like token usage.
Meanwhile, in February, LaunchDarkly announced a Snowflake Native App for “Warehouse Native Experimentation” and acquired Houseware, a provider of warehouse-native product analytics.
These developments show LaunchDarkly’s continued evolution beyond basic feature management toward becoming a comprehensive platform for software delivery that incorporates AI capabilities, experimentation, and advanced monitoring — all aimed at helping organizations ship software more safely and confidently.
And earlier this month, LaunchDarkly acquired Highlight, an open source application monitoring platform with error monitoring, logging, distributed tracing, and session replay capabilities. This acquisition strengthens its “Guarded Releases” approach.
The LaunchDarkly team.
The Future of Software Delivery
As software development accelerates with the help of AI and other tools, LaunchDarkly’s approach to controlled, measured releases becomes increasingly critical. Company CPO Vo explained that every feature is essentially “a mini business” and a chance to experiment: “Software teams work on features because they have a hypothesis about how people are going to use this feature. That’s why we built it after all, or this feature is going to improve our conversion or this feature is going to create joy or reduce churn,” she told The New Stack.
Moreover, in a LinkedIn post commemorating the company’s 10-year anniversary, Rogers reflected on their journey: “Ten years ago, LaunchDarkly’s founders, Edith Harbaugh and John Kodumal, and a small team of developers pioneered the concept of feature management, literally creating a category to address the complexities of software development and deployment. The goal was clear, to empower developers to control feature releases in real-time, mitigate the risks associated with rapid code proliferation, and enhance deployment precision.”
With increased adoption of AI and an ever-growing demand for rapid software delivery, LaunchDarkly appears positioned to continue leading this category in the years ahead.
