A broad measurement campaign was run at Volvo aiming at the evaluation of dispersion in test-based NVH characteristics of a car body and at the derivation of reference data for judging the accuracy of CAE predictions. Within this work 6, nominally identical, vehicles were tested. Tests included operational noise on Complete Vehicle (CV) level (road noise, engine noise and idling noise), NTF, VTF & Acoustic FRF measurements in CV, Trimmed Body (TB) & TB-Stripped (TBS) configurations. Additionally, modal analysis and NTF, VTF, AFRF tests were carried out on 4 BIPs of the same vehicle type. Further, limited tests were carried out on 28 vehicles of the same type. The aim of the work was to study the development of dispersion with increasing complexity of the test object, from the BIP to TB and CV. In this respect this study is unique - there has been several studies regarding the dispersion in test results but the studies that trace the dispersion between different levels of complexity of the test object are scarce if not non-existent. In parallel with this, CAE activities were carried out in order to create CAE models corresponding exactly to the test objects, for all trim levels. Even the boundary conditions were modeled with great detail. In this way, it was possible to study the development of CAE prediction accuracy with increasing complexity of the test object. Further on, several classes of measurement points' locations were studied. In the paper the dispersion results will be presented and some comments regarding prediction accuracy and its assessment will be given.