Explore Rice

7/15/2006 12:03:00 AM

New research on optimal retail pricing provides staggering results. Researchers from Rice University developed an applied statistics model to determine optimal pricing for one product category at an automotive aftermarket retailer that projected an increase of $613,000 in annual gross profits for the company. Applying the results from that one product category to the chain's other products, which number in the thousands, could potentially result in a profit increase in millions of dollars.

________________________________________________________

The study of optimal pricing boils down to how a change in price affects demand, and then using that information to determine the best price to maximize profit.

Because of scanning infrastructure and available data, most applied-statistics marketing studies have focused on optimal pricing in the grocery retail industry, leaving much to learn (for academia) and earn (for companies) in other sectors -- until now. A new study evaluating optimal pricing for a single product category at a 3,400-store automotive aftermarket retailer (called AAR in the report to maintain company confidentiality) revealed that using similar statistical models could provide more retailers with significant profit gains.

“The results projected a national rollout for the one product category to increase gross profits by [$0.6 million],” said Seethu Seetharaman, one of the study’s co-authors and an associate professor of marketing at Rice University’s Jesse H. Jones Graduate School of Management. “Multiply that by the thousands of product categories they sell, and that’s a lot of money. As we speak, they’re already setting up a second product category field experiment.”

In addition, the data, which will be published in the November 2006 “Journal of Marketing Research,” revealed other findings that had not previously been noted in grocery retail studies. For example, store demographics, such as location, size and proximity to one of the chain’s other stores, affected price sensitivity. In particular, stores described as in the North were more price-sensitive than those in the South, and stores located near other stores in the chain demonstrated less price sensitivity. Most surprising to the researchers was that AAR’s pricing was all over the place -- sometimes higher than the optimal price and sometimes lower. (Prior studies of supermarkets commonly found prices to be lower than optimal prices.)

“That the optimal prices came out sometimes higher and sometimes lower than what AAR was charging showed a very seat-of-the-pants type of decision making. They were basing their pricing on judgment calls,” Seetharaman said.

Despite the complex statistics involved, the project’s goal was straightforward: evaluate how price affects demand, using two years of price, demand and store location data for the retailer’s 3,400 stores, to determine optimal pricing for the product category at each of the stores. However, with 90 products in the product category broken into 30 subclasses, a price change of one product can affect demand for other products. The data also needed to be adjusted through a systematic estimating method, because two years’ worth of data was not enough to build the full model. To retain AAR’s competitive advantage, the product category (e.g., mufflers, brake pads, etc.) and specific pricing were kept confidential.

Seetharaman formulated the multidimensional, nonlinear model by plotting price and demand for the 30 different subclasses over an 800-store sample. Then he plugged the data into a computer program he had developed over several years of research that required several rounds of refinement to arrive at each product’s optimal price point within each of the 800 stores. He then was able to set optimal pricing for the remaining 2,600 stores by comparing store demographic information. “When I brought back the optimal pricing recommendations,” he said, “AAR said, ‘We can’t go that far.’ Some of the proposed price changes were too drastically different from what they were currently running, so we launched a smaller test run.”

On the heels of such valuable findings in a new retail sector, Seetharaman is currently working on applied statistics projects in other industries, including pharmaceutical and gasoline retailing.

Seetharaman joined Rice’s faculty in 2004 after receiving his Ph.D. from Cornell University and spending six years as an assistant and then associate professor of marketing at Washington University. He serves on the editorial board for both “The Journal of Marketing Research” and “Review of Marketing Science.”

________________________________________________________

For more information, contact Seetharaman at seethu@rice.edu or Laura Hubbard in the Jones School at lhubbard@rice.edu.