Abstract
Traditional bundling strategies often fail when production costs are high due to "overinclusion"—the inefficient sale of items to customers who value them less than their marginal cost of production. This paper introduces and analyzes Pure Bundling with Disposal for Cost (PBDC), a mechanism where customers purchase a bundle but retain the right to "return" any item for a refund equal to its production cost. We demonstrate that this simple scheme is remarkably robust, computationally efficient, and achieves near-optimal profits across diverse valuation distributions and cost structures.
The Problem: The Failure of Traditional Bundling
In monopolist pricing, bundling is typically a powerhouse strategy. By selling items together, firms leverage the law of large numbers: while individual valuations vary, the average valuation of a bundle is more predictable, allowing for greater capture of consumer surplus.
However, this logic collapses under high production costs. In a standard Pure Bundle (PB), the firm must set a high price to cover costs, which risks alienating customers. More critically, it leads to overinclusion loss, where items are sold to individuals who value them at less than their cost to produce. While Mixed Bundling (pricing every possible subset) can mitigate this, it is computationally prohibitive, requiring $2^n - 1$ different prices for $n$ items. The authors seek a mechanism that maintains the simplicity of a single bundle while achieving the efficiency of complex subset pricing.
The PBDC Mechanism
The authors propose Pure Bundling with Disposal for Cost (PBDC) as a solution. The mechanism operates in three steps:
- The firm sets a single price $P$ for the "grand bundle" of all items.
- The customer chooses whether to purchase the bundle.
- The customer can "return" any subset of items and receive a refund equal to each item's production cost ($c_i$).
Mathematically, the firm is indifferent between spending $c_i$ to produce an item or returning $c_i$ to the customer. However, this option incentivizes the customer to purchase the bundle initially, as they effectively only pay for items they value above cost. This effectively "outsources" the exclusion problem to the customer, who self-selects the optimal subset.
Theoretical and Empirical Performance
The paper provides a rigorous theoretical foundation, proving a new lower bound: either PBDC or selling items individually (Pure Components, PC) will earn at least 1/5.2 (~19.2%) of the optimal profit. This improves upon the previously established constant of 1/6.
Empirically, the authors tested PBDC against five common valuation distributions (Exponential, Logit, Lognormal, Normal, and Uniform) across scenarios involving heterogeneous items and costs.
As shown in the simulation results, PBDC is exceptionally robust. While Pure Bundling (PB) profit plummeted to as low as 16.8% of the optimum in high-cost scenarios, PBDC maintained at least 97.5% of the near-optimal profit across all tested distributions. It consistently outperformed both PC and Bundle-Size Pricing (BSP), despite BSP requiring $n$ parameters to optimize compared to PBDC's single parameter.
Welfare and Scaling Analysis
A key insight of the paper is PBDC's performance as the number of items ($n$) increases. Unlike other schemes that show diminishing returns, PBDC's profit curve is slightly convex; each additional item increases the customer's incentive to engage with the bundle.
As $n$ grows, PBDC extracts nearly 100% of the available welfare. Crucially, as illustrated in the welfare statistics, PBDC completely eliminates the Overinclusion Loss that plagues PB and BSP. For $n=6$, PB incurs an overinclusion loss of 0.356, whereas PBDC incurs none, leading to a significantly higher Producer Surplus (1.409 vs 0.553 for PB).
Conclusion and Future Work
PBDC offers a practical, near-optimal solution for firms selling multiple items with significant production costs. While the model assumes frictionless returns, the authors suggest a "checkout" version—where customers decide which items to exclude before shipping—to mitigate logistical overhead. Future research may explore optimizing the refund amount as a second parameter or relaxing the assumption of independent valuations across items.
Summary updated