Abstract
As more firms adopt big data analytics to better understand their customers and differentiate their offerings from competitors, it becomes increasingly difficult to generate strategic value from isolated and unfocused ad hoc initiatives. To attain sustainable competitive advantage from big data, firms must achieve agility in combining rich data across the organization to deploy analytics that sense and respond to customers in a dynamic environment. A key challenge in achieving this agility lies in the identification, collection, and integration of data across functional silos both within and outside the organization. Because it is infeasible to systematically integrate all available data, managers need guidance in finding which data can provide valuable and actionable insights about customers. Leveraging relationship marketing theory, we develop a framework for identifying and evaluating various sources of big data in order to create a value-justified data infrastructure that enables focused and agile deployment of advanced customer analytics. Such analytics move beyond siloed transactional customer analytics approaches of the past and incorporate a variety of rich, relationship-oriented constructs to provide actionable and valuable insights. We develop a customized kernel-based learning method to take advantage of these rich constructs and instantiate the framework in a novel prototype system that accurately predicts a variety of customer behaviors in a challenging environment, demonstrating the framework’s ability to drive significant value.
Supplemental File
Supplemental data for this article can be accessed on the publisher’s website at https://doi.org/10.1080/07421222.2018.1451957
Notes
1. All studies cited in the references for the main body of our manuscript were reviewed, as well as additional papers listed in online Appendix F (over 200 relevant studies in total). While it is not feasible to review all relevant studies, this creates a representative set from which we may draw conclusions about the completeness of our construct set.
2. We evaluated other observation period lengths (60 and 90 days), but the small improvements in predictive power provided were outweighed by the reduction in value caused by longer lead times to predictions.
3. No data for the firm and environment characteristics construct is included in our prototype system. As noted in the description of this construct within the broader framework, although potentially useful for prediction, this construct is likely more informative of shifts suggesting models be revisited as firms monitor analytics solutions after deployment.
4. We would like to thank Peter Fader for his support as we implemented this model, as well as for providing the data set used in his paper so that we could verify our implementation to be identical.
Additional information
Notes on contributors
Brent Kitchens
Brent Kitchens ([email protected]; corresponding author) is an assistant professor of information technology (IT) in the McIntire School of Commerce at the University of Virginia. He holds a Ph.D. in information systems from the Warrington College of Business, University of Florida. His research interests include customer analytics, health IT, and online information dissemination. He has published in various journals, including Information Systems Research and Decision Support Systems. Before pursuing a career in academia, he worked for five years in IT Risk Advisory at Ernst & Young LLP.
David Dobolyi
David Dobolyi ([email protected]) is a research scientist in the Center for Business Analytics at the McIntire School of Commerce, University of Virginia. He received his Ph.D. from the University of Virginia, and his primary research interests involve quantitative modeling, experimental cognitive psychology, and artificial intelligence, with applications including cybercrime and health. He has published in numerous journals, including Science, and his publications span a broad range of topics among which are reproducibility in science, eyewitness memory, Parkinson’s disease, and learning styles theory.
Jingjing Li
Jingjing Li ([email protected]) is an assistant professor of information technology in the McIntire School of Commerce at the University of Virginia. She received her Ph.D. from the Leeds School of Business, the University of Colorado at Boulder. Her research interests relate to machine learning and big data analytics, with applications in e-commerce, platform business, health care, search engine, user-generated content, and recommender systems. She received the AWS Research Grant, and Microsoft Research Azure Award for her work on big data analytics. Previously, she was a scientist at Microsoft, where she proposed and implemented large-scale machine learning solutions for Microsoft products such as Xbox One, Windows 8 Search Charm, Windows Phone App Store, Cortana, and Bing Entity Search.
Ahmed Abbasi
Ahmed Abbasi ([email protected]) is Murray Research Professor of Information Technology and director of the Center for Business Analytics in the McIntire School of Commerce at the University of Virginia. He earned his Ph.D. from the University of Arizona. His research interests relate to predictive analytics, with applications in online fraud and security, text mining, health, and customer analytics. He has published over 70 peer-reviewed articles in the leading journals and conference proceedings, including Journal of Management Information Systems, MIS Quarterly, ACM Transactions on Information Systems, IEEE Transactions on Knowledge and Data Engineering, and others. His projects on cyber security, health analytics, and social media have been funded by the National Science Foundation. He received the IBM Faculty Award, AWS Research Grant, and Microsoft Research Azure Award for his work on big data. He serves as senior editor or associate editor for several journals. His work has been featured in several media outlets.