Dynamic Pricing for Multi-Product Consumer Electronics Trade-in Program

We consider a dynamic pricing problem for consumer electronics trade-in program, where a firm buys and sells multiple types of pre-owned (used) products over a finite selling horizon. The trade-in program offers two options: trade-in for cash and trade-in for upgrade, where in the former customers sell their products to the firm and receive cash payment while in the latter they exchange their products for new products at discounted prices. The firm sets trade-in prices (both cash rewards and new products' discounts) to acquire used products and selling prices to resell them after simple refurbishments (refurbished products) to maximize its total expected profit over the selling horizon. Customer arrivals follow independent Poisson processes. We model customer preferences and choices on both trade-in and selling sides using the Multinomial Logit (MNL) model. In view of the challenge of solving the optimal prices using dynamic programming due to high dimensional state space, we develop simple and provably effective heuristic policies based on the solution to a deterministic upper bound problem. In the first policy called the Static Control (SC) policy, we set buffers at the beginning of the selling horizon to accumulate a certain amount of refurbished products inventories before starting to sell them. We show that the profit loss of this policy is in the order of O(T^{1/2}), where T is the number of selling periods. Meanwhile, we also derive a profit loss lower bound in the order of \Omega(T^{1/2}) of any static stationary policy. In the second policy called the Batched-Adjustment Control (BAC) policy, we divide the selling horizon into different consecutive and disjoint batches for different products and update the prices based on the realized uncertainties in the previous batch. The profit loss of the BAC policy is in the order of O(T^{1/3}), which is significantly smaller than that of the SC policy. Furthermore, we extend both heuristic policies to the case where the firm needs to decide the initial stocking levels of new products (for upgrade purposes). Finally, we numerically test the performance of both heuristic policies, and demonstrate that the BAC policy has superior performance over the SC policy. Some model parameters in the numerical study are estimated using real data from a major consumer electronics trade-in platform in China