Feature models are commonly used in software product line engineering as a means to document variability. Since their introduction, feature models have been extended and formalised in various ways. The majority of these extensions are variants of the original tree-based graphical notation. But over time, textual dialects have also been proposed. The “textual variability language” (TVL) was proposed to combine the advantages of both graphical and textual notations.
However, the benefits and limitations of these notations have not been empirically evaluated up to now. In this paper, we evaluate TVL with four cases from companies of different sizes and application domains. The study shows that practitioners can benefit from TVL. The participants appreciated the notation, the advantages of a textual language and considered the learning curve to be short. The study also revealed requirements for feature modelling that were not covered by TVL.