The principles of methodology of a research project in Northern Nigeria involving the collection of both farm management and nutrition data are described. This was a study of breadth rather than depth and so clear decisions were made about what was to be collected on the basis of time costs. The use of ‘registered‐non registered’, ‘continuous—one point’ and ‘data cores’ are described and discussed. From these follow the means of sample stratification using these principles. Some ideas for improving RRA are suggested.