Product Portfolio Decisions in Non-Dominated Function Spaces
Year: 2009
Editor: Chakrabarti, A.
Author: Dabbeeru,Mohan Madan; Mukerjee,Amitabha; Boob,Neha
Section: Design - Knowledge - and Product Lifecycle - Management
Page(s): 224-231
Abstract
Product portfolio optimization is the problem of providing the widest variety of functions while minimizing the product variety. Here we propose a novel approach based on functionality modeling for the product family,which can suggest portfolio decisions, building on signi cant dimensionality reductions in product variability.Function is modeled as a set of performance metrics reflecting the degree to which various user expectations are satisfied. If product variant A is better than (dominates) B on all counts, then A would be preferred; this implies that user preferences would lie among the “non-dominated” set of designs. Using a suitable multi-objective optimization algorithm, one may estimate this non-dominated set of designs, which restricts designs to a much lower-dimensional manifold in the function space. This non-dominated set which can then be clustered in an unsupervised manner, resulting in a candidate product groupings which the design team may inspect before arriving at a portfolio decision. We demonstrate the process on simple product platforms (faucets), in two scenarios: a) as an integral design, with continuous design variables, and b) as a 2-component design in which one component is available in several standard sizes. The effect of numerical stability in the process is investigated in empirically, and the conditions under which the results would scale to large dimensional spaces are also explored.
Keywords: Product Portfolio Optimization, Unsupervised Learning, Standardization