Stratigraphic correlation may be a very difficult task when only data are sparse, as it is for most subsurface applications. In this study we used synthetic boreholes extracted from numerical deltas simulated with the process-based model Delft3D. During a short workshop we organized on stratigraphic correlation the participants (Deltares and Geological survey of the Netherlands) correlated the boreholes using pencils and colours. Different groups produced different correlations of the same profiles. These differences were caused by (a) the spacing between boreholes affecting the recognition of isolated sand bodies and of dipping surfaces, (b) the multiple possible correlations, and (c) the use of a a-priori conceptual model of the stratigraphic architecture consisting of expected facies distribution and geometries. This implies that the knowledge on sedimentary system behaviour drives better geological correlation and prediction of subsurface properties. This system understanding and prediction can be enhanced using process-based models, specifically in data-sparse areas. Patterns and geometries of deltaic stratigraphy extracted from Delft3D numerical deltas can be integrated to existing geostatistical techniques and can form the basis for multiple stochastic realizations of subsurface models.