RESEARCH · RESEARCH · #2167
StoBasket paper introduces joint upsell and cross-sell optimization to maximize contribution margin
The arXiv v1 paper 'StoBasket' formulates joint stochastic basket editing, a decision-aware recommender approach that simultaneously selects replacement (upsell) and addition (cross-sell) offers and values them by expected incremental contribution margin while modeling coupled response channels. In calibrated simulations on five public basket datasets the method captures essentially all attainable offer margin (within ~0.5% in default-price settings), yields a 10.2% lift on Instacart vs. separated selection heuristics, and provides an exact optimizer that solves real baskets in a median 0.13 ms (max 111 ms).
KEY POINTS
- The arXiv v1 paper 'StoBasket' formulates joint stochastic basket editing, a decision-aware recommender approach that simultaneously selects replacement (upsell) and addition (cross-sell) offers and values them by expected incremental contribution margin while modeling coupled response channels.
- In calibrated simulations on five public basket datasets the method captures essentially all attainable offer margin (within ~0.5% in default-price settings), yields a 10.2% lift on Instacart vs.
- separated selection heuristics, and provides an exact optimizer that solves real baskets in a median 0.13 ms (max 111 ms).
WHY IT MATTERS
By coupling upsell and cross-sell decisions and optimizing for incremental contribution margin, StoBasket yields materially higher retailer profit and offers a fast exact optimizer suitable for real-time use.