The purpose of AdSOS (Ad Selection Optimization Server) is to optimally pick the best ads to serve on publisher networks to produce maximum revenue. AdSOS does this by continuously providing for each Ad the best possible predicitons of:
Each prediction has an associated confidence interval that is required by correctly deciding between competing ads to display on a given publisher site and produce maximum revenue.
The algorithms used by AdSOS to prefom these prediction are continuously updated as new data arrives. Using a process of automated updating, testing, and competition ensure the best algorithms are being used to make these critical decisions.