Purpose – To define a methodology for comparing sensor arrays for solving magnetostatic linear inverse problems. Design/methodology/approach – A singular value decomposition related projection method is used for comparing sensor arrays and we applied it to a biomagnetic inverse problem, as an example. Furthermore, a theoretical reference sensor system is introduced and used as a benchmark for the analysed sensor arrays. Findings – The method has turned out to be effective in comparing three different theoretical sensor arrays, showing the superiority of the two arrays constituted by three-axial sensors. Research limitations/implications – The method has been applied only to the case of over-determined problems. The underdetermined case will be considered in future work. Practical implications – From the applicative point of view, the illustrated methodology is useful when one has to choose between existing sensor arrays or in the design phase of a new sensor array. Originality/value – A new methodology is proposed for comparing sensor arrays. The advantage of the methodology are to take into account the regularization in the solution of the inverse problem and to be general, not depending on a particular source configuration.