Vol. 23, No. 3, March 2014, pp. 333–347 DOI 10.1111/poms.12046 ISSN 1059-1478|EISSN 1937-5956|14|2303|0333 © 2013 The Authors. Production and Operations Management published by Wiley Periodicals, Inc. on behalf of Production and Operations Management Society Clickstream Data and Inventory Management: Model and Empirical Analysis Tingliang Huang Department of Management Science and Innovation, University College London, London, WC1E 6BT, UK, [email protected] Jan A. Van Mieghem Kellogg School of Management, Northwestern University, Evanston, Illinois 60208, USA, [email protected] The copyright line in this article was changed on 11 August 2014 after online publication. e consider firms that feature their products on the Internet but take orders offline. Click and order data are disjoint W on such non-transactional websites, and their matching is error-prone. Yet, their time separation may allow the firm to react and improve its tactical planning. We introduce a dynamic decision support model that augments the classic inventory planning model with additional clickstream state variables. Using a novel data set of matched online click- stream and offline purchasing data, we identify statistically significant clickstream variables and empirically investigate the value of clickstream tracking on non-transactional websites to improve inventory management. We show that the noisy clickstream data is statistically significant to predict the propensity, amount, and timing of offline orders. A counter- factual analysis shows that using the demand information extracted from the clickstream data can reduce the inventory holding and backordering cost by 3% to 5% in our data set. Key words: click tracking; advance demand information; inventory theory and control; empirical research; dynamic pro- gramming; econometric analysis; big data History: Received: January 2012; Accepted: December 2012 by Panos Kouvelis, after 2 revisions. 1. Introduction and Related Literature extensive; see, e.g., Johnson et al. (2003), Moe and Fader (2004), Montgomery et al. (2004), Sismeiro and Recent Internet clickstream tracking technology has Bucklin (2004), Van den Poel and Buckinx (2005), generated the fast growing practice of web analytics Hui et al. (2009) and references therein. This literature and extensive ongoing research in academia. Indeed, is essentially about the marketing benefits of click- the Internet has changed the way business works by stream tracking because e-commerce websites serve providing new information and distribution channels primarily as sales channels. Clickstream tracking for both firms and customers. Customers can readily allows e-commerce firms to get accurate readings of obtain product information online without physically the efficiency of their websites, quickly usher a visitor visiting a firm. Firms can use clickstream tracking (referred to as “she” throughout the study) who is technology to see in real time who is visiting their about to purchase an item to a high-speed server, iden- websites and analyze detailed clickstreams to learn tify target visitors to show pop-up coupons, and so on. more information in advance. In contrast to e-commerce settings, we investigate Clickstream tracking allows firms to “learn about “non-transactional websites” that serve predomi- customers without asking” (Montgomery and Sriniva- nantly as a product catalog while orders are taken san 2003), but the associated academic research has offline. Many business-to-business (B2B) settings as been largely focused on online shopping and e-com- well as some business-to-consumer (B2C) settings fall merce: Montgomery (2001) shows that quantitative in this category. Specifically, this study stems from models that are commonly used in brick-and-mortar our interaction with a US manufacturer of industrial distribution channels prove to be useful in optimizing products, hereafter referred to as “the company.” The the use of clickstream data. The associated literature is company makes high-end roll-up doors that are cus- tomized for industrial and commercial buildings with This is an open access article under the terms of the regards to size, type of material, type of environment, Creative Commons Attribution License, which permits etc. The doors can go into new buildings or can use, distribution and reproduction in any medium, provided the original work is properly cited. replace older doors. Prices for a door range from the 333 Huang and Van Mieghem: Clickstream Data and Inventory Management Production and Operations Management 23(3), pp. 333–347, © 2013 The Authors. Production and Operations Management published by 334 Wiley Periodicals, Inc. on behalf of Production and Operations Management Society thousands to tens of thousands of dollars. Like many website is non-transactional. While it has been con- others, the company provides current and potential firmed in the literature that online click behavior is cor- customers with company, product, and contact related with purchasing behavior in e-commerce information on its website. However, the website is settings, it is much less clear whether such correlation non-transactional and the company sells its products persists in non-transactional settings because customers offline, either direct or through dealers. The com- do not have to visit the website to make a purchase. This pany hires the services of a web analytics firm that procedural separation reduces the predictive power of specializes in clickstream tracking to help demand web visits to forecast purchase orders if there is any sta- forecasting, procurement, and inventory planning. tistical relationship betweenthematall.Itisreported Our study focuses on the operational benefit of that e-commerce sales only account for 1.2% of all retail clickstream tracking by investigating its use as sales.1 Hence, the vast majority of commerce still is exe- advance demand information for procurement, pro- cuted offline, and thus our research setting addresses a duction, and inventory planning. We are interested in larger part of the economy beyond e-commerce. how, and to what extent, clickstream data from non- Due to the procedural separation, non-transactional transactional websites can improve demand forecast- websites provide the opportunity for firms to react. ing for inventory management. In particular, in this Clearly, in an e-commerce setting like Amazon, the setting of a B2B business with non-transactional infor- time lag between clicks and orders could be on the mational websites, we address the following research order of minutes, too short to adjust operational plans. questions: (1) How can we use clickstream data in The longer time separation between clicks and orders inventory management? This requires a tactical has an important benefit: if it exceeds the production or model that explicitly incorporates clickstream data in procurement lead time, the firm can respond to operations management. (2) How can we identify the changes in advance demand information. Matching statistically significant clickstream data and predic- supply with demand is one of the main issues for oper- tion functions (needed in the model) and improve the ations management. There is a vast body of literature demand forecast? (3) How large is the operational modeling advance demand information; see, for exam- value of using the advance demand information from ple, Hariharan and Zipkin (1995), Raman and Fisher € clickstreams to reduce inventory holding and backor- (1996), Chen (2001), Gallego and Ozer (2001, 2003), € dering costs in our setting? Ozer and Wei (2004), Tan et al. (2007), Wang and Tok- € We believe these questions are timely and impor- tay (2008), and Gayon et al. (2009). Ozer (2011) pro- tant for several reasons. The recent fast-growing vides a comprehensive literature review. All these research using clickstream data has already demon- studies assume that advance demand information is strated the great interest and importance for e-com- available and study how to use it in inventory manage- merce firms. The same applies to offline-selling firms. ment. On one hand, our study is in the same spirit of, Understanding consumer online browsing behavior and complementary to, this literature by introducing a and its value helps firms make investment decisions practical decision support model that endows classic regarding the adoption of clickstream tracking inventory management with clickstreams as a flow of technology. Manyika et al. (2011) report that “big advance demand information. On the other hand, our data—large pools of data that can be captured, com- study is the logical precedent: to what extent can municated, aggregated, stored, and analyzed—is now advance demand information be obtained from click- part of every sector and function of the global econ- streams? Although the value of advance demand omy.” Clickstream tracking has allowed individuals information is well established and understood theo- around the world to contribute to the amount of big retically, research on how advance demand informa- data available to companies. Our study examines the tion is obtained in practice and its empirical evidence potential operational value that clickstream data, an seems largely absent in the operations management lit- € important type of big data, can create for companies erature. Ozer (2011) offers several examples of obtain- and seeks to illustrate and quantify that value. In a ing advance demand information in practice such as concrete setting of the company, we show that