The Role of Machine Learning in Predicting Customer Purchase Intention
Notice: This is a sample project for study and reference. Submitting it as your own work violates most universities' academic integrity policies.
Abstract
About This Research Topic
Customer purchase intention has become one of the most closely studied outcomes in modern marketing, because knowing in advance which customers are likely to buy allows a business to spend its limited marketing budget where it will do the most good. Over the past decade, the tools used to estimate this intention have shifted away from simple surveys and scoring sheets toward machine learning systems that learn directly from browsing history, transaction records, and engagement data. This shift promises sharper targeting and better returns on marketing spend, and it sits at the centre of a wider and growing body of research on AI-driven customer analytics that many businesses now draw on to justify investment in predictive tools.
Yet a model that predicts accurately is not automatically a model that changes decisions for the better. A growing body of behavioural research shows that managers do not always trust, or act on, the recommendations an algorithm produces, even when those recommendations are demonstrably more accurate than human judgement. This article examines that gap directly, drawing on a study of digitally-enabled businesses in Enugu State, Nigeria, that investigated how machine learning capability and data quality shape the predictive accuracy of purchase intention models, how that accuracy translates into marketing decision effectiveness, and, critically, how much of that translation depends on whether managers actually trust the algorithm's output. See our broader coverage of AI-driven customer analytics in marketing decision-making for related findings on this theme.
Main Abstract
Businesses increasingly rely on machine learning (ML) to forecast whether a given customer will make a purchase, using behavioural, transactional, and demographic data to sharpen targeting and stretch marketing budgets further. Yet how far this predictive capability actually improves marketing decisions remains under-studied in emerging markets, and prior work on algorithmic decision-making warns that accuracy alone does not guarantee better decisions, since managers do not automatically trust or act on what an algorithm tells them.
This study investigated the part machine learning plays in predicting customer purchase intention among digitally-enabled businesses in Enugu State, Nigeria, focusing on how ML capability and data quality shape predictive accuracy, how predictive accuracy in turn affects marketing decision effectiveness, and how managerial trust in algorithms moderates that relationship. Grounded in the Technology Acceptance Model, Diffusion of Innovation Theory, and Algorithm Aversion/Appreciation Theory, the study used a descriptive survey design. A structured questionnaire was administered to marketing managers, data analysts, and business owners drawn from a population of 1,200 registered businesses; applying the Taro Yamane formula produced a sample of 300 respondents, selected through stratified random sampling. Data were analysed using descriptive statistics alongside Chi-square tests, Pearson correlation, and multiple regression, run in SPSS version 26.
The results showed that both ML capability and data quality significantly and positively affect predictive accuracy; that predictive accuracy significantly and positively affects marketing decision effectiveness; and that managerial algorithm trust significantly strengthens this last relationship, so that accurate predictions produce far greater improvements in marketing decisions where managers trust algorithmic output than where they do not. The study concludes that the business value of ML-based purchase intention prediction depends jointly on technical capability, data quality, and an organisation's willingness to act on what its models tell it; accuracy by itself is not enough. It recommends that businesses invest simultaneously in ML infrastructure and in building managerial trust, through explainable model design and a gradual, evidence-based approach to folding ML recommendations into everyday marketing decisions.
Keywords: machine learning, purchase intention prediction, predictive accuracy, algorithm trust, marketing decision effectiveness
Chapter One Preview
Background to the Study
Forecasting whether a customer will buy has always sat at the heart of marketing practice. For decades this forecasting leaned on customer surveys, market research, and comparatively simple statistical scoring. What has changed is the sheer volume of digital trace data now available — browsing paths, past purchases, demographic details, engagement signals — and this volume has opened the door to machine learning approaches that learn purchase patterns directly from historical customer data. Compared with traditional scoring methods, models such as decision trees, random forests, and neural networks can pick up complex, non-linear relationships across large numbers of behavioural variables, often out-predicting conventional statistical techniques.
The appeal is straightforward: a business that can identify high-intent customers more precisely can direct its marketing spend more efficiently, personalise offers more sharply, and lift conversion while cutting wasted outreach to low-intent segments, a case made in more depth in our review of predictive customer analytics and customer lifetime value. Large firms like Amazon and Netflix have built real competitive advantage around exactly this kind of predictive system, and cloud-based ML tools and CRM-embedded predictive lead scoring have since put comparable capability within reach of small and medium enterprises that have no in-house data science team.
Set against this optimism, a separate stream of research urges caution. Work on algorithmic decision-making has documented “algorithm aversion,” a tendency for decision-makers to distrust and under-use algorithmic recommendations even when those recommendations outperform human judgement, especially once they have watched the algorithm get something wrong even once. Other studies find the opposite pattern under different conditions, with people showing “algorithm appreciation” and weighting algorithmic advice more heavily than human advice. Read together, this literature suggests that the value a business actually gets from ML-based purchase intention prediction hinges not only on how accurate the model is, but on whether marketing managers trust it enough to act on it, a question that sits closer to organisational psychology than to data science, and one explored further in our coverage of AI-driven personalisation and customer purchase intention.
Frameworks such as the Technology Acceptance Model, originally developed to explain why people take up or resist new information systems, offer one lens on this trust problem, as does newer guidance from bodies such as the National Institute of Standards and Technology on what makes an AI system trustworthy in the first place.
In Nigeria, where digital transformation of marketing is moving quickly but formal data science capability remains concentrated in larger firms, it is still an open question how widely ML-based purchase intention prediction has actually been adopted, how accurate businesses perceive the resulting predictions to be, and whether that accuracy is showing up as more effective marketing decisions. This study set out to examine exactly that, among digitally-enabled businesses in Enugu State, Nigeria, with close attention to ML capability, data quality, predictive accuracy, and managerial algorithm trust.
Statement of the Problem
Nigerian businesses operate under real pressure to market efficiently in a competitive, resource-constrained environment, which is precisely what makes ML-driven purchase intention prediction attractive: the promise of sharper targeting at lower cost. Yet many businesses that have already adopted ML or AI marketing tools remain uncertain whether those tools are delivering genuine value. Part of the difficulty is that a model's technical accuracy is not something a non-technical marketing manager can directly observe or verify. Another part is that managers may simply be reluctant to substitute an algorithm's recommendation for their own judgement, particularly where the model's internal reasoning is opaque, a well-documented feature of ML “black-box” systems that has been linked directly to algorithm aversion in behavioural research.
This leaves an unaddressed empirical gap: within the Nigerian business context, it is not clear whether the technical drivers of ML predictive accuracy, namely capability and data quality, are actually being translated into better marketing decisions, or whether that translation is being blocked by low managerial trust in algorithmic output. Much of the existing literature on ML in marketing is technical, model-performance-focused research from developed markets, and it offers little insight into how model accuracy plays out in real-world marketing decisions in an emerging-market business setting, or into what role managerial trust plays in that process. This study addresses that gap directly by examining the role of machine learning in predicting customer purchase intention among businesses in Enugu State, Nigeria.
Aim and Objectives of the Study
The aim of this study is to examine the role of machine learning in predicting customer purchase intention among digitally-enabled businesses in Enugu State, Nigeria. Specifically, the study sought to:
1. Examine the effect of machine learning capability on the predictive accuracy of customer purchase intention models.
2. Assess the effect of data quality on the predictive accuracy of customer purchase intention models.
3. Determine the effect of predictive accuracy on marketing decision effectiveness.
4. Evaluate the moderating effect of managerial algorithm trust on the relationship between predictive accuracy and marketing decision effectiveness.
5. Determine the combined predictive effect of machine learning capability, data quality, predictive accuracy, and algorithm trust on marketing decision effectiveness.
Research Questions
The study was guided by the following questions:
1. How does machine learning capability affect the predictive accuracy of customer purchase intention models?
2. How does data quality affect the predictive accuracy of customer purchase intention models?
3. What effect does predictive accuracy have on marketing decision effectiveness?
4. How does managerial algorithm trust moderate the relationship between predictive accuracy and marketing decision effectiveness?
5. What is the combined predictive effect of machine learning capability, data quality, predictive accuracy, and algorithm trust on marketing decision effectiveness?
Significance of the Study
This study speaks to several audiences. For business owners and marketing managers, it clarifies whether investment in ML capability and data quality is actually paying off in more accurate purchase intention prediction, and whether managerial trust is the limiting factor stopping firms from capturing the value of those predictions. For providers of ML and AI marketing tools and CRM platforms operating in the Nigerian market, it offers evidence on adoption barriers, trust and explainability chief among them, that is directly relevant to product design. For business-support institutions and technology policymakers, it provides evidence to inform digital-capability-building programmes aimed at Nigerian SMEs. And for the wider academic community, it extends Technology Acceptance Model and Algorithm Aversion/Appreciation research into an under-studied emerging-market marketing context. Students and researchers working on related topics can find comparable methodology and structure through ScholarNestHub's research coaching service, which supports original work on marketing analytics, AI adoption, and consumer behaviour.
Scope of the Study
The content scope of the study covers machine learning capability, data quality, predictive accuracy, algorithm trust, and marketing decision effectiveness. Its geographical scope is limited to digitally-enabled registered businesses operating in Enugu State, Nigeria, spanning the retail/e-commerce, financial/fintech, telecom/tech services, and other professional services sectors. The study covers business practices and reported outcomes within the period 2023 to 2026, and relies on respondents' perceptions of predictive accuracy and decision effectiveness rather than independently audited model performance data.
Operational Definition of Terms
Machine Learning Capability: An organisation's technical and human ability to develop, deploy, or effectively use machine learning models, whether built in-house or sourced from third parties, to predict customer behaviour.
Data Quality: The accuracy, completeness, timeliness, and relevance of the data available to train and run machine learning models used for purchase intention prediction.
Predictive Accuracy: The extent to which a machine learning model's predictions of customer purchase intention are perceived to correctly reflect what customers actually go on to do.
Algorithm Trust: The degree of confidence marketing managers place in the outputs and recommendations of machine learning models, and their resulting willingness to act on them, a concept closely tied to the trustworthiness characteristics set out in NIST's AI Risk Management Framework.
Marketing Decision Effectiveness: The degree to which marketing decisions, such as targeting, budget allocation, and campaign design, informed by purchase intention predictions are perceived to improve outcomes such as conversion and campaign efficiency.
Conclusion
Machine learning has undeniably expanded what businesses can know about which customers are likely to buy, but this study's findings are a reminder that predictive power on its own does not automatically produce better marketing decisions. Capability and data quality drive accuracy; accuracy drives decision effectiveness; but the size of that final step depends heavily on whether the people making decisions actually trust what the model tells them. For Nigerian businesses investing in ML-based marketing tools, the practical lesson is to treat technical infrastructure and managerial trust-building as two halves of the same investment, not as separate projects. Readers working on related final-year research can browse more sample research projects on marketing analytics and AI adoption for further methodological reference.
Frequently Asked Questions
1. What is customer purchase intention in marketing?
Customer purchase intention refers to how likely a customer is to buy a particular product or service, usually estimated from behavioural, transactional, or demographic signals before the purchase actually happens.
2. How does machine learning predict purchase intention?
Machine learning models such as decision trees, random forests, and neural networks learn patterns from historical customer data, including browsing behaviour, past purchases, and demographics, to estimate the likelihood that a specific customer will buy.
3. Why doesn't higher predictive accuracy always improve marketing decisions?
Because marketing decisions depend on managers acting on a model's output, and research on algorithm aversion shows that managers do not always trust or use algorithmic recommendations, even accurate ones.
4. What is algorithm aversion?
Algorithm aversion is the tendency of decision-makers to distrust and avoid algorithmic recommendations after seeing the algorithm make even a single error, even when it still outperforms human judgement overall.
5. What role does data quality play in ML-based prediction?
Data quality, meaning how accurate, complete, timely, and relevant the underlying data is, directly affects how well a machine learning model can learn genuine purchase patterns, and therefore how accurate its predictions are.
6. Can small businesses use machine learning for purchase intention prediction?
Yes. Cloud-based ML tools and CRM-embedded predictive lead scoring have made basic purchase-intention prediction accessible to small and medium enterprises that do not have in-house data science teams.
7. What theories explain managerial trust in algorithms?
The Technology Acceptance Model, Diffusion of Innovation Theory, and Algorithm Aversion/Appreciation Theory are commonly used to explain why managers accept or resist algorithmic recommendations in decision-making.
8. How was this study conducted?
The study used a descriptive survey design, sampling 300 marketing managers, data analysts, and business owners from digitally-enabled businesses in Enugu State, Nigeria, and analysed the data using SPSS version 26.
9. What did the study find about managerial trust?
It found that managerial algorithm trust significantly strengthens the relationship between predictive accuracy and marketing decision effectiveness, meaning accurate predictions improve decisions far more where managers trust the algorithm.
10. What should businesses do to get more value from ML-based marketing tools?
Invest in both technical ML infrastructure and in building managerial trust, through explainable model design and a gradual, evidence-based approach to incorporating algorithmic recommendations into marketing decisions.
Purchase to unlock the full material.