Analytics for an Online Retailer: Demand forecasting and Price Optimization at Rue La La
We partnered with the online flash sales retailer Rue La La to develop and implement a pricing decision support tool that sets initial prices for new products. Rue La La is in the online fashion sample sales industry, where they offer extremely limited-time discounts (typically 2-3 days) on designer apparel and accessories. Our approach is two-fold and begins with developing a demand prediction model for new products. The two biggest challenges faced when building our demand prediction model include estimating lost sales due to stockouts, and predicting demand for styles that have no historical sales data. We use descriptive analytics (clustering) and predictive analytics (regression) to address these challenges and predict future demand. Regression trees - an intuitive, yet non-parametric regression model - prove to be effective predictors of demand.
We then formulate a price optimization model to maximize revenue of new products using demand predictions from the regression trees as inputs. In this case, the biggest challenge we face is that each style’s demand depends on the price of competing styles, which restricts us from solving a price optimization problem individually for each style and leads to an exponential number of variables in the price optimization problem. Furthermore, the non-parametric structure of regression trees makes this problem particularly difficult to solve. We employ prescriptive analytics by developing a novel reformulation of the price optimization problem and creating an efficient algorithm that allows Rue La La to optimize prices on a daily basis for the next day’s sales.
To implement our price optimization algorithm, we developed a fully-automated pricing decision support tool that runs automatically every day, providing price increase recommendations to merchants for events starting the next day. To estimate the tool’s impact, we conducted a field experiment on approximately 6,000 styles from mid-January through May 2014 to address the following two questions: (i) would implementing the tool’s recommended price increases cause a decrease in sales, and (ii) what impact would the price increases have on revenue?
For our field experiment, we used the Wilcoxon rank sum test to test the null hypothesis that raising prices according to the pricing decision support tool's recommendations has no negative impact on sales. We performed this test on styles in different price ranges, and the results suggest that raising prices only negatively impacts sales for very low-priced styles; for other price points, our results suggest that raising prices according to the tool’s recommendations does not decrease sales. Furthermore, we estimate an increase in revenue due to accepting our tool’s recommended price increases to be approximately 10% with a 90% confidence interval of [3%, 18%].
Since the conclusion of our field experiment, we have been extending our research to a dynamic pricing setting where Rue La La changes the price of a style over the course of their short selling season. As common in retailing, Rue La La has limited inventory and does not know the consumer's purchase probability at a given price and thus must learn this probability from sales data. We propose an efficient and effective dynamic pricing algorithm, which builds upon the well-known Thompson sampling algorithm used for multi-armed bandit problems by creatively incorporating inventory constraints into the model and algorithm. Our algorithm proves to have both strong theoretical performance guarantees as well as promising numerical performance results when compared to other algorithms developed for the same setting.
Our collaboration with Rue La La shows that combining machine learning (descriptive and predictive analytics) and optimization (prescriptive analytics) into a pricing decision support tool has made a positive financial impact on Rue La La's business. We hope that the success of this work motivates retailers to investigate similar techniques to help set initial prices of new products, and, more broadly, that researchers and practitioners will use a combination of analytics techniques to harness their data and use it to improve business processes.