An investigation of the structure of sizing systems

Most apparel sizing systems are based on one or two body dimensions and rely on assumptions about proportional body relationships to project other dimensions necessary to design the garment pattern. Garments from these systems will not fit a population with large variations in body proportions. Using nonlinear optimization methods and anthropometric data of US Army women three multidimensional sizing systems were derived that are designed to provide improved fit for women from the USA, a population with much variation. These systems range from an optimized linear system with a regular grade to an unconstrained optimized system with a grade break at each size. The optimized sizing systems compare favorably with D5585‐94 in their ability to accommodate variation in the population based on a mathematical test of the aggregate loss of each system. Issues related to pattern grading, size selection, and calculation of stock keeping units are discussed.