The treatment effect is a universal concept used in many fields. Regardless of the context, it is crucial to define the measure of the success of the intervention and the method of assessing the effect, which may require the use of various statistical methods. An additional challenge for analysts may be caused by imprecise data on the basis of which they are to assess the size of the treatment effect. It turns out that although the use of fuzzy sets to model imprecise data is something natural, further and deeper analysis of the treatment effect based on such data entails many difficulties. This contribution discusses how to deal with some these problems.