What regularities emerge as countries develop a pattern of built settlement? This paper uses satellite data to trace the evolution of some 50,000 built areas in Sub-Saharan Africa between 1975 to 2014, a period in which total built area increased by a factor of 2.4 due to growth and merger of settlements and the birth of new settlements. The median growth rate of settlements in the smallest initial size bin was twice that of settlements in the largest (of five) bins, rejecting Gibrat's law. Settlements of different size generally specialise in different activities, and we model this by supposing three settlement types: agricultural, agro-processing, and manufacturing/ service based. In the presence of many dispersed agricultural settlements the model predicts regular spacing of fewer and larger agro-processing settlements, and few large manufacturing/ service settlements. This pattern of spacing arises as settlements of the same type are in a competitive relationship with each other (competing for inputs and for sales of output), while settlements of different types are in a complementary relationship (because of input-output relationships). We confirm this empirically by grouping settlements into three size classes and regressing each settlement's growth on its proximity to settlements in the same and other size classes. A fast growing neighbour of similar type reduces growth, while proximity to fast growing settlements of a different type increases growth.