Determining the right pricing strategy can make or break the overall profitability of a firm. One such strategy, dynamic pricing, long practiced in the airline and hotel industries, is showing promise and profitability in the world of retail. When applied to products sold over a short sales season—new toys, skiwear, and the like— dynamic pricing can boost profits for a firm, according to research recently conducted by two professors at the Olin School of Business at Washington University in St. Louis.
Yossi Aviv, associate professor of operations and manufacturing management, and Amit Pazgal, associate professor of marketing, examined the value of dynamic pricing—the adjusting of prices based on variables such as sales history and in-stock inventory to maximize revenues in an uncertain marketplace. Their findings offer important insights for effective management of a firm’s revenues.
In their paper, titled “Pricing of Short Life-Cycles Through Active Learning,” the researchers consider a retail store that sells a quantity of a product over a fixed period of time. Their model incorporates three types of uncertainty: the frequency and timing of customer arrivals to the store, the price each customer would pay for the product, and the success of the product in the market.
“Only a small fraction of new products succeed,” Aviv said. “The sale of products such as action figures that correspond to popular movies is coordinated with release of the movie. You can’t be late in getting product out. Sales are high for a limited amount of time, then decline sharply. Often, you don’t have the opportunity to restock product. On the other hand, if sales fall short, leftover inventory is hard to get rid of. Such uncertainty can be resolved through sales observations.”
Aviv and Pazgal compare optimal expected revenues to those obtained under three different pricing practices—fixed pricing schemes, certainty equivalent policies, and pricing policies that ignore market uncertainty. This approach enabled the two to study the value of proactively setting prices to impact revenue as well as the value of continuous learning about market reaction from sales observations.
They found that a fixed-price policy, which performs very well in the case of full information about market condition, can be harmful in environments with high levels of uncertainty. For example, when a seller establishes a set price for the entire sales season, he forfeits the ability to apply lessons learned from change in market conditions.
Dynamic pricing strategies, on the other hand, become critical in settings with high uncertainty regarding market condition. In this situation, note the professors, the seller should plan for the better scenario, especially if the sales season is long. Optimal prices fall continuously over time, but at the points of sale jump upward, due to fewer units to sell and the seller’s heightened perception of the market.
In cases where the retailers suspect that consumers may feel exploited by price increases, the authors suggest an alternative model that prohibits such price increases. In all cases, since sales of short-life-cycle products are often associated with a high level of demand uncertainty, a key component of an informed sales strategy is the intelligent gathering and integration of demand information.
The researchers contend that companies have favored fixed-pricing over dynamic pricing policy because they lack dynamic pricing tools and strategy, find implementation of price changes too costly, and succumb to the call for standard pricing in a competitive market.
“One way to avoid the complexities associated with the full optimal solution to the dynamic problem is to utilize certainty equivalent heuristics,” Pazgal said. “Such heuristics separate the dynamic decision process into two parts: a passive learning component and a certainty based pricing scheme. We were surprised to find that such heuristics perform very well for most demand scenarios.”
While a large body of work on optimal dynamic pricing strategies exists in many research disciplines, a control model with a built-in learning mechanism that tracks market conditions throughout a retail sales season has never been introduced. “To our knowledge, no quantitative model for joint inventory replenishment [of non-perishable goods] and pricing decision problem appears in the literature for the case in which some parameters of the demand process are unknown,” state the researchers.
They added that their pricing decision process is easy and fast to compute and calls for easy-to-implement control strategies.
They show that obtaining market information at the beginning of the sales season is particularly beneficial when the number of units for sale is small and when the sales season is sufficiently long. Alternatively, the larger the number of items for sale, the higher the opportunity to apply the learning from actual sales observations before stock is depleted.
Aviv and Pazgal are working on several related research projects. The first develops practical algorithms for scenario-based dynamic pricing. Under this approach, retailers optimize their pricing given a set of market conditions, each specifying potential customer valuations, store visiting intensity, and market dynamics.
A different project explores ways by which retailers should dynamically price their products when faced with forward-looking consumers who time their purchases in anticipation of future discounts. “If sellers incorrectly assume that customers ignore future discounts, the sellers risk losing significant potential revenue,” Pazgal said.
Their research is funded, in part, by the business school’s Boeing Center for Technology, Information, and Manufacturing.