Analytics for an Online Retailer

Kris J. Ferreira; Bin Hong Alex Lee; David  Simchi-Levi

Article | Manufacturing & Service Operations Management | Winter 2016

Analytics for an Online Retailer: Demand Forecasting and Price Optimization

by Kris J. Ferreira, Bin Hong Alex Lee and David Simchi-Levi

Abstract

We present our work with an online retailer, Rue La La, as an example of how a retailer can use its wealth of data to optimize pricing decisions on a daily basis. Rue La La is in the online fashion sample sales industry, where they offer extremely limited-time discounts on designer apparel and accessories. One of the retailer's main challenges is pricing and predicting demand for products that it has never sold before, which account for the majority of sales and revenue. To tackle this challenge, we use machine learning techniques to estimate historical lost sales and predict future demand of new products. The nonparametric structure of our demand prediction model, along with the dependence of a product's demand on the price of competing products, pose new challenges on translating the demand forecasts into a pricing policy. We develop an algorithm to efficiently solve the subsequent multi-product price optimization that incorporates reference price effects, and we create and implement this algorithm into a pricing decision support tool for Rue La La's daily use. We conduct a field experiment and find that sales do not decrease due to implementing tool recommended price increases for medium and high price point products. Finally, we estimate an increase in revenue of the test group by approximately 9.7% with an associated 90% confidence interval of (2.3%, 17.8%).

Keywords: Online Technology; Price; Forecasting and Prediction; Revenue; Sales; Retail Industry;

Format: Print

Find at Harvard Read Now

Citation:

Ferreira, Kris J., Bin Hong Alex Lee, and David Simchi-Levi. "Analytics for an Online Retailer: Demand Forecasting and Price Optimization." Manufacturing & Service Operations Management 18, no. 1 (Winter 2016): 69–88.

The note introduces a variety of methods to assess the accuracy of machine learning prediction models. The note begins by briefly introducing machine learning, overfitting, training versus test datasets, and cross validation. The following accuracy metrics and tools are then described: mean squared error (MSE), mean absolute deviation (MAD), Brier score, and cross-entropy, true/false positives/negatives, the confusion matrix, true positive rate (sensitivity or recall), false negative rate (Type II error rate), precision, true negative rate (specificity), false positive rate (Type I error rate), receiver operating characteristic curve (ROC) and area under the curve (AUC), and precision-recall curve.

Michael Joyce, Vice President of Inventory Management at JOANN, championed an effort to develop and implement an inventory allocation analytics tool that used advanced analytics to predict in-season demand of seasonal items for each of JOANN’s nearly 900 stores and optimize store-specific inventory allocation decisions guided by these predictions. After the analytics tool had been in use for a full season of products, Joyce was surprised to hear that not everyone was pleased with its outcomes. With all of the concerns raised, should Joyce halt the use of the analytics tool for the next season?

As more commerce and media consumption are being conducted online, a wealth of new opportunities are emerging for personalized advertising. We propose a general methodology, Model Trees for Personalization (MTP), for tackling a broad class of personalized decision-making problems including personalized advertising. The MTPs learn an interpretable market segmentation driven by differences in user behavior. Using this methodology, we design two new algorithms for fundamental problems in personalized advertising – Choice Model Trees (CMTs) for the user ad-response prediction problem, and Isotonic Regression Model Trees (IRMTs) for the bid landscape forecasting problem. We provide a customizable, computationally-efficient, and open-source code base for training MTPs in Python. We train our IRMT algorithm on historical bidding data from three different ad exchanges and show that the IRMT achieves 5-29% improvement in bid landscape forecasting accuracy over the model that a leading demand-side platform (DSP) provider currently uses in production.

Faculty & Research

Analytics for an Online Retailer: Demand Forecasting and Price Optimization

Abstract

Citation:

About the Author

More from the Author