In this paper, we propose a methodology based on the use of clustering techniques derived from data analysis and multi-attribute decision analysis methods aiming at purposeful multidimensional poverty measurement. The issue of meaningfulness is thus analysed both from a theoretical point of view (measurement theory) and from an operational one (policy effectiveness). Through this new methodology of multidimensional poverty measurement, we aim at providing a contribution to methodological knowledge insisting on the necessity to build "meaningful measurements" for policy making and policy implementation. On the other hand, "meaningful measurements" appear as a contribution to the operationalisation of Sen's capabilities approach. Our standpoint underlines the necessity to consider the problem of poverty measurement as a decision problem and to tackle its measurement issue with that in mind.