arrowCase studies

Atlassian moves faster and continuously delivers software

Before

Higher mean time to recover (MTTR)

After

Improved MTTR by 97%

NPS scores increased in first year

About Atlassian

Atlassian is a well-known software company that aims to unleash the potential of every team. Best known for products such as Bitbucket, Confluence, Jira Software, Jira Service Desk, Hipchat, Statuspage and Trello, Atlassian currently serves more than 125,000 customers worldwide. Its products help teams organize, discuss, and complete shared work via project tracking, content creation and sharing, real-time communication, and service management.

Challenge

A few teams within Atlassian used an internally-built feature flagging system. It was primarily used to control feature releases, and with this system, deployments were also tied to releases. This proved sufficient at first but, but as more teams wanted to use this in-house solution, some challenges arose: the system took time and resources to maintain. Different teams working with different languages needed help developing their own Software Development Kits (SDKs). And with no interface, this system was virtually unusable for non-technical users. The result was that teams outside of engineering, including Product Management, relied on engineering support to turn features on/off or run beta tests, which became unruly when more teams wanted to use the solution.

Solution

With LaunchDarkly, more and more teams across the organization now have the ability to separate code deployments from feature releases. The first team to adopt LaunchDarkly was able to get started pretty quickly. SDKs are standardized, well maintained, and available in all major languages, so developers can setup LaunchDarkly in their stack in a matter of minutes. In the months following, other teams started using LaunchDarkly. For the first time, teams had visibility into their feature flags from a central location. When developers begin to build a feature, they simply wrap it in a flag, and then control it from the LaunchDarkly dashboard. This means teams can ship code at any time with flags turned off and features hidden, and then turn the features on when they’re ready.

Teams also have better control over segmenting their user base. Instead of switching something on/off for everyone, they now build cohorts and expose different groups to different features in a highly targeted fashion. Teams are now testing, getting feedback, and iterating while working through projects without having the changes dependent upon each other.

/ /

My teams can focus on things that are really going to add value to our customers, rather than worrying about other things that are coupled but not directly related to a specific feature. This translates to teams spending more time focusing on the items that differentiate us as a company.

Bevan Blackie

Head of Engineering - Collaboration Essentials, Atlassian

Results

The easy-to-use interface made it possible for other members of the triad or product management team to create their own specific cohorts, independently run beta tests, gather valuable feedback on new features and product usage. Once engineering puts a flag in place, anyone can easily control the flags from the LaunchDarkly dashboard. Non-technical team members started to become more involved in the feedback loop and development cycles.

Within a year, one team saw NPS scores increase—with customer feedback specific to performance and usability—and MTTR hours improve by 97 percent. These improvements have been driven by a cultural shift to a continuous delivery approach. While LaunchDarkly is not the only continuous delivery tool in use, they recognize that feature flagging and feature management are an integral part of making the continuous delivery process successful.

More case studies
Building a culture of experimentation at Stash
How Hireology enables faster, safer software delivery and migration
SmugMug increases conversions with experimentation and A/B testing