Detecting and analysing changes in consumer behaviour during life events

Darler, William (2019) Detecting and analysing changes in consumer behaviour during life events. PhD thesis, University of Nottingham.

[img] PDF (Thesis - as examined) - Repository staff only - Requires a PDF viewer such as GSview, Xpdf or Adobe Acrobat Reader
Download (11MB)

Abstract

Consumer behaviour changes over time as people experience life events, stages and transitions. There has been little empirical research that analyses these changes using real world transaction data. This thesis contributes to the research on the consumer life course through the creation of two methods to predict and describe consumption behaviour during a life event. Transaction data sets from two multinational retailers are used to develop methods to 1. Detect the points in time when consumption behaviour is likely to change during a life event and predict the event-stage of new customers 2. Describe and visualise the trajectories of different customer segments over time based on the combination of items they put in their shopping baskets. These methods are demonstrated using retail data from 3.4 million customers experiencing the event of parenthood over a period of 10 years.

This thesis is split into six case studies that utilise predictive and descriptive branches of analytics to make methodological contributions to literature on consumer behaviour during life events. The first four case studies are used to develop a novel framework to divide a life event into stages based on changes in consumption behaviour and predict the event stage of new customers. In previous studies on consumption behaviour during parenthood, stages were described by biological changes (e.g. trimesters) or estimated by marketing managers with domain-specific knowledge. The new framework allows event stages to be detected without any domain specific knowledge and identifies the important products for event-stage prediction. It was found that parenthood consisted of up to seven event stages which vary between customer demographics. It is shown that these event stages conflict with biological changes and expert predictions. The final two case studies are used to demonstrate a new method to describe and visualise the trajectories of customers during a life event. This addresses the differences in consumption patterns revealed by the first study, and the paucity of segmentation techniques that show how customer behaviour varies over time. It was found that there was greater heterogeneity at customer experiencing parenthood at the retailer than at the supermarket and the new method provided richer information about customer dynamics over time than previous, static segmentation techniques.

Item Type: Thesis (University of Nottingham only) (PhD)
Supervisors: Goulding, James
Roberts, Deborah
Smith, Andrew
Keywords: Life course, Life events, Consumer behaviour, Marketing, Machine Learning, Data science, Retail, Big Data
Subjects: H Social sciences > HF Commerce
Q Science > QA Mathematics > QA299 Analysis
Faculties/Schools: UK Campuses > Faculty of Science > School of Computer Science
Item ID: 56630
Depositing User: Darler, William
Date Deposited: 18 Mar 2020 15:40
Last Modified: 06 May 2020 09:51
URI: http://eprints.nottingham.ac.uk/id/eprint/56630

Actions (Archive Staff Only)

Edit View Edit View