Whenever many large technology have a new feature they want to test, they roll it out to a small set of users before letting it go live. That those companies get feedback on the feature and iron out any bugs, and also roll back the changes quickly if anything is wrong or if the users don’t like it. But that requires teams of people to maintain those tools, and smaller startups don’t really have the resources to build them.
Now there’s a startup called LaunchDarkly that’s bringing those tools to other companies to do just that. The company is coming out of beta later this month, and said it has raised $2.6 million in order to invest more in its tools.
Here’s how it works: whenever a company tests a new feature, the code for both the original version of the tech and the new feature are both actually live. With LaunchDarkly, a company can activate the new code for only a small set of users to test new features before launching.
The company launched in 2014 when Edith Harbaugh had a couple meetings with developers, including one who had just left Amazon. That former Amazon director was upset that he didn’t have the same tools that he had at Amazon, and it was a problem Harbaugh was very familiar with from her days at TripIt and other companies. For example, she dealt with a situation where a new feature was launched and it broke another part of the service, and each subsequent attempt at fixing the broken part resulted in a new broken part.
“I’ve been in engineering for 15 years, and I’ve been part of some epicly bad releases,” she said. “I’ve seen so many releases go so bad for so many different reasons. You can never really know how your product is really gonna work until it’s actually out in the hands of customers. You can do usability studies, quality assurance, but it’s your users that are always gonna give you the feedback. That was the realization I made — if you can get actual features out to people sooner, you can get better and better feedback.”
Her co-founder John Kodumal — who went to school with Harbaugh — happened to have a Ph.D. in computer science and statistics, and felt that he knew how to build the tools. So they decided to spin it into a company and have enlisted investors like SoftTech VC in order to make it a reality.
The surprise to the pair, she said, was that it was a tool that larger companies were also interested in using beyond just startups. While other companies have used it as a competitive edge, it had been something that was lacking at technology companies of all sizes, meaning that there was a large market for those kinds of tools — and a massive opportunity for a company to come in and build a business.
That’s because if a launch goes poorly — and many often do — it can not only alienate a user base, but also potential customers. And if too many launches go poorly, it can inadvertently drive a service, or even a business, into the ground. Many startups have become popular simply on the basis of being easier to use than other kinds of software that have poor user experiences. And tools like LaunchDarkly help startups and technology companies refine their products more quickly to fit the needs and wants of users.
“We’re separating out the business logic from the actual code,” Harbaugh said. “You don’t have to hack behavior any more; you have a way to control it. You can turn a feature on and off for an entire country, which is a pretty standard best practice — if you launch a feature in Canada first, and it goes well, you launch it in North America.”
LaunchDarkly, like other software as a service companies, uses a subscription model for its tools. Developers get a dashboard that allows them to get real-time feedback and toggle soft launches with small sets of users. The tools apply to massive user bases or even small ones — where only 20 users end up testing the product. LaunchDarkly then integrates with other services that measure the performance and success of those new features in order to help those companies decide which ones are doing well.
For now, LaunchDarkly is focusing on commerce and SaaS companies, but a tool like this has pretty obvious implications in other parts of the tech universe.
One great example is video games, where companies often roll out massive updates and patches and then proceed to hotfix things along the way, and run beta servers that allow users to opt-in to test those features. It would allow game developers to tweak small things and test them with a live user group — not a group that is tolerant of bugs and opts in to beta tests — and figure out which ones work.