Aurifeille, J.Medlin, C.2006-06-192006-06-192001European Journal of Economic and Social Systems, 2001; 15(2):3-161292-88951292-8909http://hdl.handle.net/2440/1185In business science, the studied objects are often groups of partners rather than independent firms. Extending classical segmentation to these polyads raises conceptual problems, principally: defining what should be considered as common or specific at the partners' and at the segment levels. The general approaches consist either in merging partners characteristics and performances into a single macro-object, thus loosing their specific contributions to each partner's performance, or in modelling partners' performance as if their models were independent. As a step to understanding, how partnership influences firms' performance, the dyadic (i.e. two partners') case is studied. First, some theoretical issues concerning the degrees of individual and contributive interest in a dyadic population are discussed. Next, partnership's conceptualisation is based upon two models for each firm: a "self-model" that reflects how the firm's characteristics explain its own performance, and a "contributive-model" model that reflects how these characteristics influence the partner's performance. This allows definition of three relationship modes: merging, teaming and sharing. Subsequently, dyad segmentation strategies are discussed according to their capacity to reflect the modes of partnership and a dyadic clusterwise regression method, based on a genetic algorithm, is presented. Finally, the method is illustrated empirically using actual data of business partners in the software market.en© EDP Sciences 2001Business partnershiprelationshipssegmentationdyadsgenetic algorithmJournal article002002142510.1051/ejess:200111259826Medlin, C. [0000-0003-0567-2538]