New trends in the management of manufacturing efficiency are associated with the development of such areas of Data Science as Big Data and Data Analytics, taking into account the human factor. With asymmetric awareness of management and personnel about the potential for efficiency, undesirable activity of personnel is possible. In this case, personnel as the active element selects its state in such a way as to maximize its own objective function. For this reason the energy efficiency of manufacturing can be not equal to the potential. Therefore, traditional methods and algorithms of Data Science may be ineffective. To solve this problem in the face of uncertainty, adaptive decision support system with coacher's instructions is proposed. In this system, the manager is considered a disciple of the coach, even if both of them are not aware of the potential of the energy efficiency. In order to ensure the effectiveness of traditional learning algorithm of Data Science, that system includes rating and stimulation procedures. Sufficient conditions have been found for the synthesis of the decision support system which is learning with coacher's instructions, in which the active element uses the potential of energy efficiency. The use of such system is illustrated by the example of energy efficiency of a wagon repair company