Field experiments are a key source of data and knowledge in agricultural research. An emerging practice is to compile the measurements and results of these experiments (rather than the results of publications, as in meta-analysis) into global datasets. Our aim in the present study was to provide several methodological paths related to the design of global datasets. We considered 37 field experiments as the use case for designing a global dataset and illustrated how tidying and disseminating the data are the first steps towards open science practices. We developed a method to identify complete factorial designs within global datasets using tools from graph theory. We discuss the position of global datasets in the continuum between data and knowledge, compared to other approaches such as meta-analysis. We advocate using global datasets more widely in agricultural research.