Abstract: A system and method for advertisement exposure forecasting are described herein, which advantageously provide viewer-level advertisement exposure forecasting for linear media platforms. Based on historical viewing data, a viewing proportion estimation model is trained for each of a plurality of viewers/devices. A plurality of viewing proportions is determined with respect to a plurality of future programs for each of the plurality of viewers/devices. Based on the plurality of viewing proportions, a reach and frequency of an advertising campaign are determined based on advertisement spot counts for each future program that is involved in the advertising campaign.