Abstract

Recommendation systems are central to electronic commerce, yet most models overlook how user preferences evolve along the conversion funnel. This study introduces a lightweight, funnel-aware add-on that reorders baseline recommendations according to a user’s stage in the purchasing process. The approach is designed as a general, easily deployable layer on top of any recommender, offering clear implementation and reproducibility. Using two publicly available e-commerce datasets from cosmetics and electronics domains, we demonstrate two additive improvements: first, that a simple similarity-based reranking substantially outperforms a raw popularity baseline; and second, that incorporating funnel-stage awareness provides an additional performance lift. Even in extremely sparse environments where collaborative filtering struggles, the method improves hit rate and recall through stage-aware personalization. These findings highlight the potential of aligning recommender outputs with marketing funnel theory to enhance both relevance and conversion in online retail.

Download

Share

COinS