122/264 We are currently developing a proprietary artificial intelligence technology that we believe will have even more applications leveraging our data. We believe this technology will optimize self­serve promotional inventory and content promotion scheduling. Leveraging Our Unique Data Assets We own a large and diversified data set which provides us with significant insight into content consumption and User behavior. As of December 31, 2017, this data set is more than 200 petabytes (compared to 60 petabytes at Netflix as of November 2016 according to publicly available data) and approximately five petabytes are queried daily (compared to approximately three petabytes queried daily at Netflix as of November 2016 according to publicly available data). Since all our Users have to log in to access our Service, this enables us to track behaviors such as playing songs, sharing, selecting recommended music, skipping, following, and active participation through the upvote and downvote buttons. More than 150 billion of these kinds of events are logged daily on our Service. Furthermore, we devote substantial resources to analyzing and handling this data in order to obtain useful insights with machine learning and artificial intelligence. We have the ability to personalize and curate the content we stream by measuring an individual User’s preferences against more than 40 different parameters. Our algorithms are designed to anticipate a User’s preferences using factors such as demographics and past listening behavior. Furthermore, we can combine situational context, such as time of day and location, to make better recommendations for appropriate content to an individual User based on his or her current activity. As we further collect and process data and understand our Users, we believe our technology will better understand and respond to our Users’ preferences, drive an even better overall User experience, and further differentiate our Service from our competitors. Our Technology Architecture We have designed our technology architecture with a number of core principles in mind to ensure that it is scalable, multi­platform, easy to update, and built to support future development. • Scalable, Cloud­Based Infrastructure. We maintain a capital­light infrastructure. In 2016, we began to migrate over to the GCP for the majority of our computing, storage, bandwidth, and other needs. By using a cloud services provider, we are able to focus on running and scaling our Service rather than building our own infrastructure. Given the capabilities of our partner, we can continue to meet our Users’ needs as we grow. As of December 31, 2017, the majority of our computing needs were serviced from our cloud infrastructure. • Mobile First. Our mobile architecture is the product of years of development experience, and we continue to improve and adjust its features. Our mobile interface is built to be outwardly sleek and aesthetically pleasing, while capturing the full power and utility of our platform. It has been developed for those on the go, integrating the latest advances in human­machine interaction for touch screen fluidity. • Platform Agnostic . We believe that Users should be able to enjoy their music wherever, whenever, on any device. We continue to grow the number of our key partnerships with hardware vendors. We support OS X, iOS, Android, Chrome OS, Fire OS, Windows, Windows Phone, BlackBerry OS, Linux and Symbian. • Rapid Updates and Agile Development. Through our advanced development environment, we are able to quickly distribute product improvements, constantly bringing an improved Spotify experience everywhere. We employ agile development techniques and processes, making continuous and flexible product improvements. 115