Aggregate resource assessments, derived from three subsequent generations of voxel models, were compared in a qualitative way to illustrate and discuss modelling progress. We compared the models in terms of both methodology and usability. All three models were produced by the Geological Survey of the Netherlands. Aggregate is granular mineral material used in building and construction, and in this case consists of sand and gravel. On each occasion ever-increasing computer power allowed us to model at a higher resolution and use more geological information to constrain interpolations. The two oldest models, built in 2005 and 2007, were created specifically for aggregate resource assessments, the first as proof of concept, the second for an online resource information system. The third model was derived from the ongoing multipurpose systematic 3D modelling programme GeoTOP. We used a study area of 40 × 40 km located in the central Netherlands, which encompasses a section of the Rhine-Meuse delta and adjacent glacial terrains to the north. Aggregate resource assessments rely on the extent to which the occurrence and grain size of sand and gravel are resolved, and on proper representation of clay and peat layers (overburden and intercalations) that affect exploitability. Average model properties (e.g. total aggregate content) are about the same in all three models, except for a difference resulting from converting older lithological classifications to the current one. This difference illustrates that data selection and preparation are paramount, especially when dealing with quality issues. Generally speaking the results of the aggregate resource assessments are consistent and satisfactory for all three models, provided that they are judged at the appropriate scale. However, the assessments based on GeoTOP best approach the desired scale of use for the aggregates industry; in that sense progress was significant and each model was a better fit for the purpose.