An online system receives information describing actions performed by individuals and identifies online system users included among these individuals. Based on the actions they have performed, the users are assigned to sequentially ordered stages of a classification scheme associated with a content-providing user of the online system. The online system receives content items associated with different stages from the content-providing user, in which content items associated with a stage may be presented to users assigned to the stage. The online system may determine an expected return value associated with each stage and an expected advancement value associated with advancement of users assigned to each stage to succeeding stages of the classification scheme. The online system also may determine a value associated with a content item based on a comprehensive value received from the content-providing user, the expected advancement value, and a predicted likelihood of user advancement to a succeeding stage.