Extended conditional trend analysis: Predicting triple period buyer flows with a tri‐variate NBD model. Issue 1 (22nd July 2021)

• Record Type: Journal Article
• Authors: Trinh, Giang; Dawes, John; Wright, Malcolm J.; Danenberg, Nick; Sharp, Byron
• Other Names: Anesbury Zachary William guestEditor.; Stocchi Lara guestEditor.; Trinh Giang guestEditor.
• Is Part Of: Journal of consumer behaviour. Volume 21:Issue 1(2022)
• Page Start: 92
• Page End: 101
• Publication Date: 2021-07-22
• Subjects: Consumer behavior -- Periodicals; Motivation research (Marketing) -- Periodicals; 658.834205
• DOI: 10.1002/cb.1979
• Languages: English
• ISSNs: 1472-0817

Abstract

Abstract: Marketers are interested in the loyalty of their customer base. Increasingly this includes examining behavioural loyalty inferred from the frequency or weight of purchase. A typical approach is to divide the customer base into arbitrary segments based on weight of purchase and then attempt to move customers from lighter to heavier segments, rather than have them reduce purchasing or cease buying altogether. Effects can be monitored by examining how purchasing by groups of individuals evolves over successive periods. However, much of the flow between segments represents random fluctuations in period‐to‐period purchasing rather than true change to underlying loyalty. Accurate analysis requires true change to be separated from these stochastic changes, for example through benchmarks derived from conditional trend analysis (CTA). While CTA considers the two‐period case, it provides no guidance for changes seen across three‐periods. The three‐period case is nonetheless regularly reported by panel companies and relied on by managers. We therefore develop the three‐period CTA, using a tri‐variate NBD, to allow the analysis of buyer flow across three successive periods. We provide an empirical illustration and demonstrate fresh insights into the evolution of consumer loyalty. The findings allay oft‐raised concerns about supposedly 'lost' buyers, as perceived customer loss is often simply regression to the mean of the buying rates. Accordingly, the three‐period CTA shows predictable proportions of buyers who move between different buying‐weight segments, including first‐year buyers who were apparently 'lost' in the second year but return to buy the brand in the third year.