Customisation with crossed-basis sub-segmentation

Abstract

The paper proposes a cost-effective method of creating customised marketing contacts, and tests the method with a large field experiment. The firm should sub-segment its customers in different ways using complementary basis variables, and create marketing contacts that are comprised of content determined by the various sub-segments. A field experiment illustrates and evaluates our approach, showing that customisation improves response rates, but not basket sizes. The paper also tests the 'labelling hypothesis' for newer customers, which asserts that the firm is better off guessing (labelling) the segment membership of a newer customer (about whom little is known because of limited transaction history) than sending a generic offer. We find that customised offers outperform generic offers for newer customers.