A new formal strategy in the multidimensional optimisation of the experimental variables affecting the chemiluminescence (CL) detection in flow injection analysis (FIA) is proposed here. The strategy implies several steps, being the most significant: selection of the variables to be studied and their experimental domain; use of a screening design to detect significant variables and interactions into the experimental region; study of the main effect of variables and second-order interactions; and finally application of a Draper-Lin small composite design (orthogonal) to obtain the optimum values of the significant variables. The methodology is applied to the determination of methylamine by FIA based on the use of the peroxyoxalate CL (PO-CL) reaction. Considering the high number of experiments required due to the different chemical and instrumental variables to be taken account and their adequate compatibility to obtain maximum sensitivity, the methodology offers a rigorous study of the main effects and interactions, achieving a reduction of experimental work.