Abstract: Technologies are described for adaptive selection of user to database mapping. An impact to processing resources by a user of a database managed by the processing resources may be predicted using criteria based on user-database properties such as a number of rows in database, a number of rows in database for the user, a maximum depth of hierarchy for the user, a maximum breadth of hierarchy for the user, applications that have recently accessed the user's data, distribution of quota types in the user's data, a number of shared items in the user's hierarchy, a number of people who share in the user's shared item hierarchy, an application employed by the user to access the data currently, etc. Based on a model for the predicted impact, a behavior change (for any of the criteria) and/or a move of the user to another database may be suggested/automatically implemented.