Techniques are described for generating statistics that reflect the consumption of media resources, such as television program resources. The techniques involve polling a group of devices identified in a random sample list or a custom sample list to collect event data describing the consumption of resources by the devices. Lower-bound and upper-bound thresholds are applied to remove or modify event data that may not accurately reflect the legitimate consumption of resources. Varying probability weights can also be assigned to events to reflect their likelihood of representing the legitimate consumption of resources. Based on such filtered event data, for a given device, a computation technique computes the percentage of time that the user has consumed a particular resource relative to a total amount of time in a time segment. Similar percentages are computed for other devices and other time segments to provide plural percentage values. Rating information and share information are computed based on these calculated percentage values. The tracking functionality compiles the calculated statistics into various web-enabled UI presentations and provides these reports to an operator over an intranet or like network.