Interpretable machine learning models for customer satisfaction: A comparison of decision trees in the coffee shop context
Interpretable machine learning models for customer satisfaction: A comparison of decision trees in the coffee shop context

Summary/Abstract: Research background: Customer satisfaction is a critical determinant of competitive performance in service markets, and the coffee shop segment is highly competitive, and experience driven. Companies increasingly predict satisfaction from survey data using machine learning, yet the most accurate models are often opaque, while managers need insight they can understand and act on. Purpose of the article: The purpose is to classify overall customer satisfaction among coffee shop customers using interpretable models and to identify the service attributes that most strongly distinguish satisfied from dissatisfied customers. The study draws on the view that satisfaction reflects the perceived performance of the service rather than customer characteristics. Methods: The study uses primary data gathered through an online questionnaire completed by 400 customers of a global coffeehouse chain between October 2025 and February 2026. Four decision tree algorithms were compared, the training set was balanced by oversampling the minority group, and the models were evaluated on a separate test set using accuracy and sensitivity. The most accurate model was interpreted through the importance of its predictors and the structure of its tree. Findings & Value added: All four models classified satisfaction with accuracy above 93%, and the best model reached almost 95% while reliably identifying dissatisfied customers. Satisfaction was driven entirely by assessments of service attributes rather than customer characteristics, and the cleanliness and appearance of the interior was by far the strongest driver, followed by product quality. The study shows that interpretable models can achieve high accuracy while offering managers actionable guidance on satisfaction drivers.

• Details
• Contents