Discount Sensitivity–Are We All Equal?

Asnat Greenstein-Messica, Lior Rokach, Asaf Shabtai

The high adoption of smart mobile devices among consumers is an opportunity for ecommerce retailers to increase their sales by recommending consumers with real time, personalized coupons which take into account the specific contextual situation of the consumer [5]. While context-aware recommender systems (CARS) have been widely analyzed [1, 2] personalized pricing or discount optimization in recommender systems to improve recommendations’ accuracy and commercial KPIs has hardly been researched [7]. This paper studies how to model user-item personalized discount sensitivity and incorporate it into a real time contextual recommender system in a way which can be integrated into a commercial service. We propose a novel approach for modeling user-item personalized discount sensitivity in a sparse data scenario, and present a new CARS algorithm (CBRF) which combines co-clustering [3, 4] and random forest [6] to incorporate the personalized discount sensitivity.We conducted an experimental study with real consumers and mobile discount coupons to evaluate our solution. In this experience, less attractive coupons were offered with higher discount. We defined two features, which model the user and item independent discount sensitivity: 𝑑𝑢𝑢 the probability of a user 𝑢 to consume a coupon with a discount level 𝑑, and 𝑑𝑖𝑖 the probability of an item 𝑖 to be consumed utilizing a coupon with discount level 𝑑. The new features are expressed by