The value of market demand information in serial supply chains with Markov -modulated demand

This research quantifies the value of market demand information for serial supply chains facing Markov-modulated demand. Companies facing this type of demand may or may not know the demand distribution in every time period, since shifts in the demand pattern happen probabilistically. Higher uncertainty about the demand distribution leads to higher inventory-related costs. To reduce these costs the companies may be inclined to conduct market research to learn the demand distribution, thereby reducing demand uncertainty and related costs. We quantify the cost savings resulting from such market research by comparing the average cost of the optimal inventory policy under partially observed demand with the average cost of the optimal inventory policy under fully observed demand for a single location and for a two-echelon serial supply chain. To calculate the average cost of the optimal policy under the partially observed Markov modulated demand the discrete approximation is used. We prove that the average cost of the serial supply chain can be found by adding retailer's average cost to the supplier's average cost. The retailer's average cost is obtained by solving the retailer's single location problem assuming perfect supply. The supplier's average cost is obtained by solving the supplier's single location problem with induced penalty cost added to the supplier's one time period cost. The induced penalty cost is charged when the system-wide inventory is insufficient to replenish the retailer's inventory to the optimal level. It represents an increase in the retailer's expected cost under insufficient supply. The results indicate that the value of market demand information increases as the difference between the demand distributions in the different states increases and as the autocorrelation of the demand process increases. The value of market demand information decreases as the lead times in the supply chain increase.