Modèle paramétrique des coûts de production des principaux produits d'Amos

Abstract

AMOS S.A. is a Belgian medium sized company specialized in manufacturing of high precision optical, mechanical and opto-mechanical systems for the space industry and professional astronomy. Each project undertaken by the company is quite unique, this heterogeneity makes the quotation process long and costly. Therefore, it is difficult for AMOS to quickly deliver quotes to potential customers. To overcome this issue, the company has decided to create a parametric estimate cost model for its main products, namely telescopes and mirrors. Such model is built on mathematical relationships, called “CERs”, between some of the main technical characteristics of the studied product and its costs, using regression models. This paper presents the development of a parametric model for AMOS’ mirrors. All the necessary steps of the development are described. Moreover, the diverse theoretical concepts linked to the mathematical aspect of this project are explained in detail. Emphasis is put on the use of General-Error Regression Model (GERM) to develop a more accurate model than one obtained with Ordinary least squares (OLS) regression. During the development, it became obvious that the current cost traceability of AMOS doesn’t always allow to identify the unit cost of each mirror within a specific project, hence making the real cost of the company not convenient for the model development. Therefore, the model relied on the cost found in the company’s internal quotations. However, these are estimates and imply unavoidable errors. This study results in the demonstration of the current impossibility of developing an accurate parametric model for AMOS. A review of some potential solutions is included at the end of this paper to help the company solving this issue.