Over the past decades, many studies have focused on the dimensions of channel-belt sand bodies (referred to as "channel belts") in fluvial deposits because of their relevance for hydrocarbon exploration and production. Some field studies have revealed a significant downstream decrease of channel-belt width and width/thickness ratio along the length of the channel belt. To verify whether this is a common feature in fluviodeltaic settings, eight Holocene channel belts in the Rhine–Meuse delta (The Netherlands) were selected and one in the Lower Mississippi Valley (U.S.A.). We determined channel-belt geometry (width or width/thickness ratio) using geological–geomorphological maps and detailed cross sections based on borings. It was found that the width of channel belts encased in cohesive deposits decreases by a factor of 4 to 6.5 in a downstream direction along the length of the channel belts. The width/thickness ratio decreases by a factor of 2.5 to 5. On the other hand, the width of a channel belt incised in a noncohesive substrate remains constant along its entire course. These observations are related to longitudinal changes in bank erodibility and stream power. It is suggested that bank erodibility is the dominating factor determining the geometric properties of channel belts, and the variability therein, at least in the Rhine–Meuse delta. The currently available alluvial-architecture models most likely overestimate sand quantities and connectedness in alluvial successions, because channel-belt dimensions are held constant in all models. We therefore propose that the factors "bank erodibility" and "stream power," which influence width and width/thickness ratio of channel belts, should be incorporated in future alluvial-architecture models in order to make more realistic estimates of sand quantities in river deltas.