In 2016, Pandora chose SingleStore as the data store and foundation for solving a specific business problem. The goal was to build a dashboard that would allow their Ad sales team to view data about current and former ad campaigns in real-time. The previous system had limitations such as a two-day delay for new data to be available, and performance issues due to pre-computing fixed answers for specific queries. Pandora required a data store that could deliver results via real-time queries, support two years of raw click and impression data, and be fast enough to support sub-second query response over hundreds of billions of rows of data. The contenders included RedShift, CitusDB, and SingleStore, with SingleStore offering the right mix of features, performance, and flexibility to address the problems Pandora wanted to solve. Pandora worked closely with the SingleStore team to implement the solution, building a test cluster and eventually deploying it in production. They found that columnar data provided significant benefits for query performance, but required careful consideration of indexes, data types, and hardware configurations. The implementation also involved addressing issues such as OR vs IN clauses, data skew, and execution plans. Pandora ultimately chose SingleStore due to its ability to meet their project requirements, features, and performance needs, which have remained true even after expanding their usage beyond the initial Ads dashboard project.