The assessment of the soil resource of any region has two parts, namely, an inventory of the kinds of soil and their distribution, and knowledge of the way each kind can be used and its performance under a range of circumstances. Soil varies substantially and intricately over short distances in most parts of the world. Inventory by field survey and air-photo interpretation must be done at a local scale. Inventories may be combined so that an individual nation state or region of similar size can know what kinds of soil it has, how much and where they are, how much each can produce, how to manage each in perpetuity, and the risks of degradation in use. Local classifications, with classes defined simply and identifiably on aerial photographs, will serve for mapping, and in combination with classical statistics can provide sound estimates from stratified sampling and agronomic experimentation.Sound assessment should also be at this local scale initially. This should combine fundamental understanding of the soil's behaviour, strategic agronomic research on regional stations, and on-farm trials. The last are crucial for estimating productivity of the soil in practice.Data from all sources can be stored, sorted and displayed by geographic information systems that now have abundant capacity. They should be indexed by soil class and other attributes, with clear distinction being made between assessments of productive potential and basic data. They should be publicly accessible, to ensure that data are readily available and never lost.Estimates of the soil resource and its productivity for large regions, nation states, and the world can be compiled from local surveys by sampling through a ' bottom-up ' procedure.