Contemporary trends in customer profitability studies have led to the awareness that product costs are only one part of the business-customer relationship costs. This relationship is burdened with the various costs generated by business activities that support this relationship. Allocation of these costs leads to the conclusion that buyers, in accordance with their characteristics, differ greatly in the size of the associated costs and that each monetary unit of the individual customer's income does not participate equally in the company's profits. The diversity of aforementioned costs and its allocation to individual customers is enabled by the introduction of new cost accounting methods. Based on these methods, a model of calculation of actual customer profitability indicators has been developed. The model identifies a set of variables for internal and external environment that are important for calculating revenue, costs, and profits per each customer. The outcomes of the model are profitability measures of the individual customer. The empirical value of these indicators is the basis for taking management actions focused on managing the customers based on past data. However, the management goal is not only reactive but proactive acting onto the customer database. Proactive action is based on the results of the model that holds even predictive capabilities. The basis for predictive ability of the model was found in the application of techniques and methods of multivariate statistical analysis and machine learning. The input variables of the model are also non-financial, in addition to purely financial ones, which are by their nature mostly unsuitable for processing using traditional statistical methods. Methods of Multivariate Statistical Analysis exceed some of these deficiencies and may be partially used to analyze data on customer profitability. Machine learning methods can reveal hidden knowledge and patterns of customer behavior from retrospective analyzes by taking into consideration nonlinear and nonfinancial elements in addition to linear variables. The ability to approximately describe any uninterrupted function makes them a good choice of methods for predicting customer profitability. The aim of this paper is to build a model for measuring current and predicting the future profitability of an individual customer by using machine learning methods and Multivariate Statistical Analysis. In that process, the paper compares selected methods of machine learning and multivariate analysis with the purpose of identifying more suitable methods for predicting customer profitability.