When Edith Harbaugh and John Kodumal first met in an advanced math class at Harvey Mudd, one of the country’s top engineering colleges, neither planned on one day launching a multimillion-dollar tech company together. Especially one that would change the way developers deploy code.
“I just wanted to solve my math problems back then,” says Harbaugh, who is now the CEO of LaunchDarkly, a company that helps software developers reduce risks and prevent technical slip-ups from being exposed to the public eye.
Today, seven years after LaunchDarkly was founded, the Oakland, California-based startup has a $3 billion valuation after closing on $200 million in series D financing led by Lead Edge Capital, a growth-equity investment fund that focuses on software and tech services. Other investors pitching in the round include TTCP, Insight Venture Partners, Bessemer Venture Partners, Vertex Ventures US, Redpoint, Threshold, Uncork Capital and Bloomberg Beta.
With the announcement of its newest funding, LaunchDarkly has tripled its valuation from its previous funding round in January 2020 when it reached unicorn status. The feature management platform has a team of 300 engineers and enables more than 2,000 customers including IBM, Atlassian and TrueCar to release more features at a faster rate.
After graduation, Kodumal and Harbaugh stayed in touch and, as they progressed separately in their engineering careers, both encountered a common snag in the software creation process. “It’s launch day, and you just roll it (code) out to everybody and if something bad happens, it can be catastrophic,” says Harbaugh.“You’re just dead in the water.”
A developer’s dilemma
For developers who spend weeks, months, sometimes years, writing and testing new pieces of code for software features, it all boils down to the much-anticipated, stress-soaked launch day. If a developer deploys an erroneous code, it could bring the company website to a grinding halt and shut-off the entire customer base from using the application.
The risk of failure is high and so is the cost. Such was the case for Slack, which experienced an outage during the first working day of 2021 and then again in May. The workplace communication platform was down for a couple of hours each time. Previous outages in 2019 cost the company $8.2 million in credits to customers. In 2019, Stripe, which processes payments for millions of companies, suffered two hours of downtime, costing the company every transaction fee it would have earned if it had been up and running. Knight Capital Group survived one of the most notorious of website crashes when, in 2012 the financial services company lost $460 million in 45 minutes, almost pushing it to bankruptcy. On average the cost of downtime is $5,600 per minute, according to a study by Gartner.
For design collaboration tool InVision, one of LaunchDarkly’s early customers, crashing production was an expected part of the process. “We used to joke on our team, when the team was a lot smaller, that you weren’t really a true member of the team until you crashed production,” says Ben Nadel, cofounder of InVision, who has also been a developer since 1999. In addition to financial costs, the company also lost customer approval and development time or “butts-in-seats” time as Nadel calls it.
While working as a product director for travel app Tripit, Harbaugh wielded a tool she was closely familiar with: feature flagging, a technique of breaking down a feature release into smaller components and launching it in stages before it’s made visible to users.
“So instead of, for example, pushing out a release with 20 features, you can push out a release with one feature,” she tells Forbes.
Feature flags are like on-and-off toggles used for deploying new or risky sections of code. In other words, it’s a way to dark launch. It gives DevOps teams control over the software development process by allowing developers to decide which users, as well as how many, can see or use a new feature. For instance, instead of introducing a new feature to everyone, it can be teased to a small percentage of users or users in one specific region, testing the waters while the main operation continues to run.
While feature flagging is a common practice in software development, feature management, which allows collaboration across different silos of a company, is a new tool. The next step for Harbaugh was to fill in the gaps by joining forces with her former classmate Kodumal and starting their own company.
Harbaugh spent the initial years educating prospective customers and selling them on the perks of feature management. From Norway to Australia, she gave talks in different parts of the world. “It was incredibly hard to get those first 10 customers on board. It took a lot of stress, persuasion, and making sure that they really have confidence in us to get them to us,” says Harbaugh, who, as a cross-country cyclist and ultramarathon runner, knows the importance of endurance.
In 2020, the Covid-19 pandemic lit up a need for digital experiences and online components across all industries. More companies moved to the cloud and digital went from being a nice-to-have to a must-have, Harbaugh says. As a result, five-year business roadmaps for some LaunchDarkly clients collapsed into two to three weeks, she says. On peak days LaunchDarkly helps companies ranging from restaurant delivery to garment manufacturers deploy 20 trillion feature flags.
A carpentry challenge
Companies that attempt their own in-house feature management technology rather than outsourcing it face a process similar to carpenters attempting to build their own hammers, Harbaugh explains—an arduous task she took on as one of her college engineering projects.
It took three tries—and a lot of wasted wood and metal —for her to get it right.
“After I finished building this hammer, I never wanted to build another hammer again,” she remembers. “I think with developers, there’s this weird thing where if you show up at a job site as a carpenter, and the client said, ‘you need to build a hammer and a screwdriver before you start work,’ they’ll be like, ‘well, I’m never gonna get anything done if I have to build my own tools.’”
Indeed, it’s these very companies that decide to build their own tools which end up being the biggest preachers of LaunchDarkly’s services, bringing in more customers, according to the cofounder.
Though LaunchDarkly still has a majority of funding left from its previous round, the fresh cash will help accelerate its growth, scale to more customers and hire 100 new engineers. It also plans to continue to build integrations with other services such as ServiceNow, Allasian, Zendesk and Salesforce.
Nadel, who has been using feature flags for the past six years, describes the addition as a paradigm shift. Allowing engineers to program and release features incrementally has decreased the massive “cognitive overhead” associated with big projects, he says. Though it was slow to adopt it, the company has created nearly 1,500 feature flags to date.
“The beauty about feature flags is it allows you to react to things that you didn’t expect,” he says. “If it’s behind a feature flag… then you have a lot more leeway in what you feel empowered to do as an engineer.”