Animoto delivers software 3X faster with LaunchDarkly

Animoto's award-winning online video maker makes it easy for anyone to drag and drop their way to powerful and professional marketing videos. The company’s certified partnerships with Facebook, Instagram, YouTube, Twitter, and the Small Business Administration give it unique insight into the changing social media and business landscape, which may be why more than 1 million businesses around the world have used Animoto to create marketing videos that stand out on social media and beyond. Founded in 2006, Animoto has offices in New York City and San Francisco.

Competition in the video creation software market has intensified. In the face of this pressure, Animoto resolved to find a way to release new features more frequently and drive higher levels of user engagement. For this to happen, the company's approach to software delivery had to change.

Before LaunchDarkly, Animoto built and tested new features in a staging environment. Product managers, designers, and testers reviewed features in staging and gave feedback to engineers, who then reworked the underlying code, put it back in staging, awaited more feedback, and so on. The longer this went on, the more challenging it became to merge new code with the master application. Not only that, it deferred the inevitable bug fixes engineers would have to make once the feature was set loose in the wild. Lastly, under this model, product managers struggled to validate new functionality in a real-world context.

Feature flags could remove these development roadblocks, at least in theory. And the engineering team did, in fact, start incorporating them. But they relied on a homegrown system for managing flags. Adding further complexity, they employed different feature flagging systems for the web and mobile versions of their application – systems that did not speak to each other. While such a solution may hold up when managing a few flags, it collapses when the few become many. Beyond the scaling dilemma, engineers also were limited in what they could do with individual flags. Ultimately, the VP of Engineering felt wary of testing in production with the homegrown system. Barriers to faster delivery thus remained.

LaunchDarkly's feature management platform gave Animoto a way to easily govern all its feature flags across all its applications – all in one place. What's more, it expanded the breadth and sophistication of feature flag use cases available to engineers – e.g., the ability to use multivariate flags versus basic Boolean types. With this foundation, Animoto could exert far greater control over software releases.

Animoto can now split users into different cohorts and perform targeted rollouts based on that segmentation. As such, engineers dictate who sees what when. This, along with LaunchDarkly's other safeguards, gave Animoto the confidence to finally test in production. The engineering team can progressively release new features, starting first with low-risk cohorts and then expanding to riskier audiences. What's more, product managers can run early beta tests in a live setting. The insights from these trial runs have helped improve the overall product quality.

Another reason why Animoto feels empowered to test in production is, in the event a feature goes haywire, an engineer (or anyone) can quickly shut it off using a kill switch. And with LaunchDarkly's streaming architecture, that change registers in 200 milliseconds. Now, if a third-party service connected to Animoto's application goes down, for example, the company can deactivate that service with one click, shielding customers from any fallout.

Animoto also conducts A/B tests with LaunchDarkly feature flags. For example, the product team measures how well different pricing and packaging options perform against one another and then pushes that data to their analytics system. This enables leadership to make better data-driven decisions that help the bottom line.

The control and operational fail-safes of the LaunchDarkly platform gave Animoto the justification it needed to shed its staging environment. In short order, the company has advanced into doing continuous integration and continuous delivery. Developers safely merge new (often incomplete) code with the master at a regular clip. Indeed, the ability to deploy code without releasing it to users – the cardinal benefit of feature flags – made this possible.

Animoto achieved its main goal of delivering high-impact features faster. Before LaunchDarkly, it took as long as three weeks to take a feature from idea to production. The engineering team cut that down to one week and is on a path to further reducing it. To borrow language from the State of DevOps report, Animoto shrunk its lead time for changes by 66%.

Animoto also reported that customers are ecstatic with some of the new functionality. As a shining example, the company recently used LaunchDarkly to iterate on a new "photo burst" feature. When the final version went live, it generated a lot of buzz among Animoto's Facebook community and garnered high praise from users.

More broadly, Animoto has vastly reduced its operational risk when delivering software – a lasting and transformative benefit. No doubt, Animoto is poised to shape the future of video creation software.