08-10: Mohamed Zaki
Chair: Tuck Siong Chung
I Love You and I Won’t Leave You”: Unpacking Customer Loyalty
Customer loyalty is a key strategic priority for organizations. It is argued that the greater the loyalty to a company, the more beneficial those customers are to the company. Accordingly, customer loyalty is considered a critical indicator of a company’s performance (Gupta et al. 2006, Oliver 1999). It is surprising then that so many firms rely on simplified, single-metric ways of measuring customer loyalty such as Net Promoter Score (NPS) and are unaware of what their customers really think of them. While such measures are easy to administer and provide a set of numbers that can be presented to the board, they fail to take into account the many dimensions to customer loyalty which cannot be measured by a single data point (Aksoy 2013; Huang and Rust 2013; Kandampully, Zhang, and Bilgihan 2015; Rust and Huang 2014, Zaki and Neely 2018, McColl-Kennedy et al. 2019).
Clearly, better conceptualization and better measurement is required (Rust and Huang’s 2004, Ostrom et al.’s 2015, McColl-Kennedy et al. 2015, McColl-Kennedy et al. 2019). To this end, the purpose of our work is to: (1) conceptualize customer loyalty as a multidimensional construct offering a novel conceptual framework that integrates prior research to unpack loyalty; (2) employ a machine learning model using multi-data sources (attitudinal and behavioral) to empirically test our conceptual framework that combines quantitative and qualitative measures over two years for a B2B service provider; and (3) provide a step-by-step guide for implementing the machine learning approach in practice, assisting managers to develop a much richer view of customer loyalty.
We employ both quantitative and qualitative measures which contains in total around 5000 records. Quantitative measures (quality, repurchase overall satisfaction, NPS, repurchase) are used as input to our machine learning model. We employ a linguistic text mining approach introduced by McColl-Kennedy et al. (2019) to analyze qualitative measures to determine the complaint status of each customer segmenting customers into four groups: (1) complaints (2) compliments (3) neutral or (4) providing suggestions. Customer transaction survey data is then transformed into profitability scores, facilitating the categorization of customers based on their purchasing behavior. Further, dimensionality reduction and clustering techniques are used to identify customer loyalty segments. Finally, our prediction model (using a supervised classifier algorithm) correctly predicted 80% of customers likely to be loyal, based on the results of the testing dataset. Our work contributes in two important ways. First, we provide a theoretically derived conceptual framework and second, demonstrate the strengths of this new approach relative to traditional averaged scores. Finally, we offer practical insights and lessons for management.