Austin Byrne, Senior Manager
Atlassian was founded on $10,000 in credit card debt in 2002. The rapid organic adoption of tools such as JIRA and Confluence, with no sales teams, was an industry first. Millions of visitors navigated to Atlassian web properties each month, with no attribution or weblogs, which limited deep understanding of paid online marketing efficacy and the optimization opportunities that come with that knowledge.
In 2015, the marketing growth measurement team was able to begin capturing row-level weblog data via third party instrumentation, which laid the foundation for deep optimization opportunities in subsequent years. Basic dashboarding, monthly insights, and a marketing playbook were the first data feedback mechanisms for Atlassian marketers to optimize their strategy and tactics.
In 2016, building on the high fidelity historical data captured since weblog instrumentation, the team built five proprietary attribution models to better understand the efficacy of paid marketing tactics, such as display or paid search advertisements. For the first time in Atlassian history, marketers were able to dynamically and programmatically adjust their spend tactics and see results within days instead of months.
In 2017, more segmentation data was added to the models, and the addition of bid management and impression platforms scaled efforts further. Higher fidelity analytics and the growth of segmentation feature breadth has laid the foundation for a new era of optimization, this time employing machine learning and predictive techniques to scale and optimize even further.
Atlassian paid marketing reduced cost-per-trial by 50 percent and more than tripled return on advertising spend from paid search channels. The company has recovered hundreds of thousands of dollars in refunds from industry-leading vendors whose targeting algorithms didn’t behave as designed, based on Atlassian’s geographic analysis insights. Hundreds of thousands of dollars have been reallocated away from normally always-on channels that bring in significant trials, but do not convert those trials into monthly active users.