Chapter One — 1.1 Background to the Study

Retail in Nigeria's large cities has changed considerably over the past two decades. International and national supermarket chains have expanded into Lagos, convenience formats have multiplied, and online grocery platforms now deliver to many neighbourhoods. Alongside these sit a large number of family-owned supermarkets, often founded by one entrepreneur and run with the help of spouses, children or close relatives, which depend heavily on a loyal base of nearby households. For these stores, a customer who quietly moves their weekly shopping to a competitor represents a real and recurring loss, and replacing that customer usually costs more than keeping them would have.

Customer churn prediction is well established in subscription industries such as telecommunications, banking and insurance, where a customer's departure is a recorded event and historical churn labels are easy to obtain. Grocery retail is different. Shoppers are non-contractual, they split their spending across several stores, and they rarely announce that they are leaving. Research on non-contractual churn has shown that transaction histories can still reveal declining engagement, through falling visit frequency, longer gaps between purchases and shrinking basket values, well before a customer disappears entirely. Many family-owned supermarkets in Lagos already hold this kind of data in their point-of-sale and loyalty records, but they generally lack the analytical capacity, staff and tools to turn it into timely, practical decisions about which customers to approach and how.

1.2 Statement of the Problem

Family-owned supermarkets in Lagos generate a continuous stream of customer transaction data, yet churn is typically recognised only after customers have already left, and retention efforts, where they exist, are general promotions applied to everyone rather than targeted at the customers most at risk. Existing churn prediction research has concentrated on contractual industries and on large firms with dedicated analytics teams, and the systems described in that literature assume data infrastructure, technical staff and budgets that a family-run store does not have. There is little work on how churn should be defined and predicted in a non-contractual Nigerian supermarket setting, and even less on packaging such predictions into a decision support tool that a store owner can actually use without specialist knowledge. This study addresses that gap by designing, building and evaluating a predictive analytics-based decision support system for churn management tailored to family-owned supermarkets.

1.3 Objectives of the Study

1. To develop an operational definition of customer churn suited to non-contractual supermarket retail, based on inactivity and declining engagement in customer transaction histories.
2. To identify the transactional and behavioural features that best signal churn risk among customers of family-owned supermarkets in Lagos.
3. To build and compare machine learning models for churn prediction and select the best-performing model using precision, recall, F1-score and AUC-ROC.
4. To design and implement a web-based decision support system that presents churn risk scores, the main drivers behind each score, and recommended retention actions.
5. To evaluate the decision support system in terms of predictive performance and its perceived usefulness and ease of use among owners and managers of the participating supermarkets.

1.4 Research Questions

1. How can customer churn be appropriately defined for non-contractual supermarket customers using transaction data?
2. Which transactional and behavioural features are most predictive of churn among customers of family-owned supermarkets in Lagos?
3. Which machine learning model gives the most accurate and reliable churn predictions on this data, and how do the candidate models compare?
4. How can churn predictions be presented within a decision support system so that store owners and managers can act on them?
5. How do owners and managers of the participating supermarkets rate the usefulness and ease of use of the decision support system?

1.5 Significance of the Study

For the participating supermarkets and similar family-run retailers, the study offers a practical tool for spotting customers who are drifting away early enough to do something about it, and for directing limited marketing budgets at the customers where retention effort is most likely to pay off. For researchers, it extends churn prediction into non-contractual grocery retail in a West African context, where both the data characteristics and the business constraints differ from the telecoms and banking settings that dominate the literature, and it shows how a predictive model can be embedded in a decision support system designed around the realities of small business users rather than corporate analytics departments. The study may also be of interest to point-of-sale vendors and retail technology providers looking to build analytics features for the small and medium-sized retail segment in Nigeria.

1.6 Scope of the Study

The study will be limited to selected family-owned supermarkets in Lagos, Nigeria. The participating businesses will serve as the empirical case study for the design, development and evaluation of the proposed decision support system. The study will focus on customer transaction and behavioural data relevant to churn prediction, drawn from point-of-sale and loyalty records that can be linked to identifiable repeat customers, and will not attempt to model all aspects of customer relationship management, such as customer acquisition, pricing strategy, supplier management or the full cost-benefit of individual retention campaigns. Customer data will be anonymised before analysis in line with the Nigeria Data Protection Act 2023. The system will be developed and evaluated as a working prototype within the participating stores rather than as a commercially deployed product, and findings will be interpreted with regard to the specific supermarkets studied rather than generalised to all Nigerian retailers.