OPTIMIZATION MODEL SUITS FASHION RETAILER
Zara, a Spanish clothing chain, uses analytical model to manage inventory in over a thousand stores
LOS ANGELES - As a second year doctoral student at MIT, Felipe Caro began having regular discussions with Jeremie Gallien, an associate professor at the Sloan School of Management, about the unique supply chain challenges faced by the Spanish clothing retailer Zara. The discussions became more frequent and Gallien eventually became Caro's advisor.
"Zara is well-known in the operations management field," says Caro, who is now an assistant professor at UCLA Anderson. "They represent what is called Fast-Fashion. Zara can move an item from design to the shelves in four to six weeks while traditional retailers can take six or nine months."
That speed allows Zara to introduce more products during the selling season. The company is reported to stock some 11,000 items during a season compared with just a few thousand items at most stores. Zara takes items off the shelves just as fast. If a key size, typically medium, sells out, the item goes to the back room so that customers are not disappointed to find their size missing.
Caro says there was a gap in the operations management literature since most research in the field dealt with traditional apparel retailers whose long lead times force them to guess what will be selling many months later. "With reduced lead times," says Caro. "Zara has real-time sales information they can use to modify the product assortment. If they see that certain colors are selling better, or customers are buying a round neck instead of a v-neck, they can quickly make adjustments."
So Caro set about developing a mathematical model that uses sales data and other inputs to determine the optimal assortment of products to place on a store's shelves. The project resulted in his doctoral thesis. "My thesis was about what we called the 'dynamic assortment problem,'" continues Caro, "and I was very interested in seeing if it would actually work in practice. I established contact with a Zara representative named Jose Antonio Ramos and sent him my thesis."
Caro, who is from Chile, quickly received a reply in his native language inviting him to make a presentation to company executives and the distribution team. So he and Professor Gallien from MIT traveled to Spain in July, 2005. "Zara has information systems and procedures to ship garments,"says Caro, "but they had never used optimization models. We showed them the impact of these models in other industries and described the possible benefits for them."
Zara executives told Caro that their immediate need was to determine the quantities of items to be shipped each week from the firm's two main warehouses in Spain to the network of stores. "They were very good at deciding what items to put on the shelves," he says, "but they were facing more urgent challenges in terms of quantities. So, given the products in the warehouses, we started looking into all the details that come into play and finally developed an inventory allocation model."
The Zara allocation model was developed after Caro became an assistant professor at UCLA Anderson. During this time, Mr. Ramos continued to serve as Caro's liaison with Zara and as the champion of the project.
Implementation involved coding the model and linking it to databases containing sales data, orders from store managers, warehouse inventory and a variety of other inputs. Much of this work was done by a student intern named Juan Correa who became a key contributor to the project. "Models like this are very objective," says Caro. "In this case, our model maximized the sales of the entire network and not just the sales of a specific store. It looks for the global optimum rather than having each store manager trying to optimize locally."
At first, the completed model was used in just one of the main warehouses to allocate a few representative products. The other warehouse was unchanged to serve as a control. "Then, using a well-established statistical methodology," says Caro, "we were able to document a three to four percent sales increase using our model with almost no additional costs." For Zara, this is pure profit that goes directly to the bottom line.
Based on these results, Zara executives decided to implement the model at a full scale. This took place in early 2007. Now, every item in all Zara stores is shipped using this model.
Caro believes the model offers great potential for many other retailers -- particularly those trying to make complex decisions in short time frames. But for now, there is more work to be done at Zara where executives have asked Caro and Gallien to develop a mathematical model to determine the timing and magnitude of price markdowns during clearance sales. "That's a work in progress," he says, "but we hope to have a model by the end of this year."
Click here to read Felipe Caro's paper, "Inventory Management of a Fast-Fashion Retail Network."