Predicting sales success with machine learning models

Thumbnail Image
Journal Title
Journal ISSN
Volume Title
School of Business | Master's thesis
Degree programme
Information and Service Management (ISM)
Sales success is a critical aspect of business performance, influencing the profitability and sustainability of companies. Thus, predicting success gives essential insights to decision-makers by aiding with more efficient resource allocation and guiding in developing tailored sales strategies. Machine learning models have become increasingly useful tools in sales success prediction, and their importance has been noted in the manufacturing industry where machine learning capabilities are anticipated to open revenue-generating opportunities and competitive advantage in the following years. This master´s thesis is a single-company case study that explores the application of machine learning models for predicting sales success within the manufacturing service industry and discovers the key features influencing sales success within the case company. Additionally, the study investigates the interrelationships between these features from a sales success perspective. The research employs logistic regression, decision tree, XGBoost, and CatBoost models, with hyperparameter tuning applied to enhance the performance of XGBoost, to predict the sales success and find the most influential factors. Results indicate that ensemble machine learning models outperform traditional methods, with tuned XGBoost performing the best in sales success prediction. The direct sales channel has a clear positive effect on sales success and has the most positive correlations with different services and customer regions in won sales frequencies. Furthermore, factors such as technical support and repairs, maintenance, and service agreement services as well as customers in Latin America and the Caribbean and bidding units in Asia exhibit a positive impact on sales success. Conversely, replacement services, bidding units in Northern America, and the distributor sales channel demonstrate negative impacts. This case study contributes to the literature by providing a foundation for broader findings not only regarding the usability of machine learning models but also concerning the factors influencing sales success and their interrelationships in the manufacturing service industry. Moreover, the study gives decision-makers strategic insights and tools to enhance sales performance and overall business outcomes.
Thesis advisor
Vilkkumaa, Eeva
sales success prediction, machine learning, XGBoost, CatBoost
Other note