TL;DR
This paper introduces a novel strategy for manufacturers to optimize capacity planning by strategically using advance selling. By collecting customer commitments before making a final capacity decision, companies can significantly reduce demand uncertainty and improve profits, especially in volatile markets. The research provides a framework for determining the optimal timing to stop collecting these advance sales and make the capacity investment.
What Problem Does This Paper Address?
Capacity investments, like building a semiconductor fab or power plant, are notoriously expensive, irreversible, and made under significant demand uncertainty. Traditionally, companies might overbuild to meet potential demand, risking underutilization, or underbuild, missing out on sales. This paper tackles the challenge of making these crucial capacity decisions more effectively by leveraging information gathered before the investment is locked in. The core problem is: how can a manufacturer use advance selling to gain valuable insights into future demand and make a more informed, profitable capacity decision, and crucially, when should they stop gathering this information and commit to a capacity level?
Methods
The authors develop a dynamic programming model to analyze the manufacturer's decision-making process. Key elements include:
- Advance Selling as Information Acquisition: The core idea is that offering products for sale before the regular season provides a "market signal" – commitments from customers that reveal information about potential demand.
- Optimal Stopping Policy: The paper establishes the optimality of a "control-band policy." This means the manufacturer monitors cumulative advance sales. If the sales fall within a specific range (the control band), it's optimal to stop advance selling and decide on capacity. If sales are too low, continue selling to gather more information; if too high, it might be optimal to continue selling to potentially capture more revenue at a higher price before committing to capacity.
- Pricing Strategies: The model considers two scenarios:
- Exogenous Pricing: Prices are fixed in advance.
- Optimal Pricing: The manufacturer determines both advance selling and regular season prices optimally.
- Demand Model: A price-sensitive, stochastic demand model with an iso-elastic form is used, allowing for analysis of how demand changes with price and uncertainty.
Datasets
Not discussed in the paper. The study relies on a numerical simulation framework to explore various scenarios and parameter settings.
Key Insights and Ideas
- Advance Selling is More Than Just Early Revenue: While early revenue is a benefit, the primary value of advance selling lies in its ability to reduce demand uncertainty and inform capacity decisions.
- The "Control Band" is Key: The optimal strategy isn't to always stop selling as soon as possible, nor to sell indefinitely. A specific range of commitments signals the right time to act.
- Market Volatility Rewards Information: In markets with high demand uncertainty, the value of acquiring information through advance selling increases significantly.
- Capacity Costs Matter: Higher capacity building costs make advance selling more attractive, as it helps avoid costly over- or under-investment.
- Customer Behavior Influences Strategy: When customers anticipate shortages (high "late purchase tendency"), advance selling becomes more beneficial.
Key Results
- Significant Profit Improvement: The study demonstrates that advance selling can significantly improve a manufacturer's expected profit compared to a "build-it-and-they-will-come" strategy.
- Optimal Stopping Policy is Robust: The control-band policy is shown to be optimal for both exogenous and endogenous pricing scenarios.
- Conditions for Maximum Value: Advance selling is most valuable when demand uncertainty is high, customers anticipate shortages, capacity costs are high, and customer price sensitivity is relatively low.
- Dynamic Pricing Enhances Strategy: Allowing dynamic pricing during the regular sales season, in addition to advance selling, further boosts profits, though the marginal benefit of multiple pricing opportunities diminishes.
Limitations and Future Work
- Model Simplifications: The model assumes a single product and a single capacity investment opportunity. Future work could explore multi-product scenarios, flexible capacity, and product substitution.
- Demand Learning: While the market signal function can be linked to Bayesian learning, the paper doesn't explicitly model full Bayesian demand learning.
- Transaction Costs: The cost of implementing dynamic pricing or frequent price changes is not explicitly modeled.
- Future Research Directions: The authors suggest exploring the impact of advance selling with multiple products, flexible capacity, and product substitution. They also note the potential for more complex demand learning models and the integration of advance selling with after-capacity-set strategies.
Summary updated