Software product lines are typically developed using model-based approaches. Models are used to guide and automate key activities such as the derivation of products. The verification of product line models is thus essential to ensure the consistency of the derived products. While many authors have proposed approaches for verifying feature models there is so far no such approach for decision models. We discuss challenges of analyzing and verifying decision-oriented DOPLER variability models. The manual verification of these models is an error-prone, tedious, and sometimes infeasible task. We present a preliminary approach that converts DOPLER variability models into constraint programs to support their verification. We assess the feasibility of our approach by identifying defects in two existing variability models.