Implementations described and claimed herein provide systems and methods for managing inventory usage to predict when a pool of available inventory will become exhausted based on past usage trend data. In one implementation, raw usage data is extracted from a first data sample. The first data sample has a first input time, and the raw usage data specifies a total amount of inventory for one or more inventory types. The total amount of inventory is categorized for each of the inventory types into an amount of used inventory and an amount of available inventory. A projected inventory usage for each of the inventory types is generated based on the categorized total amount of inventory. The projected inventory usage defines an availability status of each of the inventory types.