We present an approach for studying three quality issues related to survey data, namely the detection of interviewer misbehaviour, the correct implementation of the questionnaire programming and possible redundant paths, which lead to case numbers hindering feasible statistical analysis. The idea is using a theoretical graph model for describing all possible paths through a survey questionnaire. The vertices of the questionnaire graphsrepresent questions and the edges the paths linking the questions. So, we are able to derive properties of the questionnaire by studying its graph. These properties are all traversed paths through the questionnaire and all possible data patterns resulting through filtering. Examining the properties gives first hints if there are peculiarities, which we need to examine in more detail. To study potential interviewer misbehavior, we test if the mean number of follow-up questions of interviewers is significantly smaller than the overall mean number of follow-up questions of all interviewers. To examine path redundancy and the questionnaire programming we test, if the path frequencies fall below a pre-specified value and compare the vertices and edges of the theoretical and empirical graph, respectively. To study the capability of our approach, we designed a small case study and applied it to NEPS data.