A method for managing a cost of ownership of a portfolio of N software applications (N.gtoreq.2). A regression function is identified for each application. Each regression function expresses an approximate cost of ownership of the applications in terms of parameters and coefficients. J applications A.sub.1, A.sub.2, . . . , A.sub.J are selected from the N applications such that a remaining N-J applications are unselected (J0). .SIGMA..sub.j represents a summation over j from j=1 to j=J. W.sub.j are predetermined weights.