Byline: Denise Power

CHICAGO — When a stand-alone technology solution delivers a 700 percent return on investment — as Gymboree’s price optimization system has done — the next logical step is to incorporate that crucial data into operational systems.
Ed Wong, senior vice president of planning, allocation, logistics and information systems at the children’s apparel retailer, said results from its analytics-driven merchandise markdown process have been so effective at guiding pricing activities that the chain will feed the resulting data into its merchandise planning systems. The company hopes to complete this phase by the end of July.
“We’ve seen a seven-time return on investment. The amount of money we made back was seven times the amount of money we spent” since implementing the system two years ago, he said.
Currently, Gymboree receives markdown recommendations — involving both the optimum timing and depth of markdowns — from a technology vendor that hosts the application on its own systems infrastructure. Technology Strategy, Cambridge, Mass., uses the retailer’s current and historical sales data and inventory information along with its own mathematical algorithms to calculate the best markdown strategy for specific product types to squeeze out the most margin.
By incorporating this data into its planning systems (from IBM, Armonk, N.Y.) Gymboree will more fully — and proactively — leverage the data that is already guiding the markdown process effectively.
Too often, retailers say, the markdown process is subject to emotions, often that of buyers who are reluctant to discount merchandise until it’s clearly not moving. At that point, however, it’s often too late and the opportunity to eke out the maximum profit has slipped away
Wong said two years of Gymboree’s historical sales data, along with a year of paper-archived data, were gathered before the price optimization project was initiated in 1998. The vendor then evaluated the historical sales patterns of different categories of merchandise and translated those patterns into mathematical algorithms for use in simulation models.
The predictive modeling exercises take into account a great number of variables, including seasonality differences from year to year and the timing of previous markdowns — “a whole slew of metrics,” said Wong.
“The engine does not just rely on history and expect the future to repeat itself,” he said. Instead, the system takes into account sales history as one factor, but also up-to-date data to identify the most profitable outcome of specific pricing actions.
“This is not someone’s guess,” Wong said. “It’s based on sophisticated mathematical algorithms.”