The importance of online customer reviews for the success of products and services has been recognized in both research and practice. Therefore, the ability to explain and interpret customer assessments expressed by the assigned overall star ratings is an important and interesting research field. Existing approaches for explaining the overall star ratings, however, often do not address methodical issues associated with these ratings (e.g., ordinal scale). Moreover, they often ignore the review texts which contain valuable information on the customers’ assessments of different aspects of the rated items (e.g., price or quality). To contribute to both research gaps, we propose a generalized ordered probit model using aspect-based sentiments as independent variables to explain the overall star ratings of online customer reviews. For measuring the explanatory power of our model, we suggest a likelihood-based pseudo R-squared measure. By evaluating our approach using a large real-world dataset of restaurant reviews we show, that, in contrast to other regression models, the generalized ordered probit model can address the methodical issues associated with the star ratings. Moreover, the evaluation shows that the results of the proposed model are easy to interpret and valuable for analysing customer assessments.