Chapter 10. Predictive approaches to customer loyalty: the impact of missing data on the predictability of customer loyalty models

Current predictive models in research on customer loyalty that rely on large amounts of observations face a huge challenge - real data is incomplete. The author argues against default approaches available in commercial statistical software packages that employ ad-hoc procedures to ignore or replace missing data. The author explains why such methods of handling missing data are unnecessary and can be potentially harmful. Based on a theoretical framework of the three types of 'missingness', the author empirically demonstrates how different methods of handling missing observations can lead to incorrect, inconsistent, or even misleading findings in empirical research on customer loyalty.