| CPC G06Q 30/0243 (2013.01) | 20 Claims |
|
1. A method for modeling incremental sales for a retail enterprise, the method comprising:
receiving guest profile data including historical campaign data for a plurality of guests and a set of features for each guest;
receiving a proposed promotion with a campaign goal;
generating a proxy campaign by:
determining an eligible subset of guests from the plurality of guests for the proxy campaign based on the proposed promotion;
for each eligible guest in the eligible subset of guests, calculating a ratio of promotional spend per item purchased by the eligible guest; and
selecting a first group of guests for a treatment segment and a second group of guests for a control segment based on the calculated ratios of promotional spend per item purchased to reflect a promotional spend per item bias in a historical campaign; and
training an uplift model to predict an uplift score for each of the plurality of guests with the proposed promotion using the guest profile data for the first group of guests in the treatment segment and the guest profile data for the second group of guests in the control segment, wherein predicted uplift scores are indicative of which guests would likely meet the campaign goal if given the proposed promotion and are unlikely to meet the campaign goal if not given the proposed promotion, and
wherein the proxy campaign is different from historical campaigns stored in the historical campaign data.
|