The human foot is a complex 3D structure, and foot shape is an important consideration in footwear design. This thesis presents a novel approach for improving footwear fit through the development of a 3D statistical shape model (SSM) of the human foot and the application of unsupervised machine learning techniques to determine fit options for footwear development. Using a dataset of over 20,000 weight-bearing 3D foot scans collected globally, the research aimed to capture population-level foot shape variation and translate it into practical, size-specific footwear fit solutions. After aligning and registering the scans into point correspondence, principal component analysis was used to identify dominant modes of shape variation. These modes revealed meaningful covariation in foot geometry, particularly in regions critical to footwear design such as the instep, heel, and forefoot. Clustering techniques were then applied within foot length classes to generate representative fit shapes, providing a data-driven alternative to the traditional “one fit per size” approach in footwear. Compared to average foot shapes, these clustered fits offered improved geometric matching across the dataset. The findings confirm that even within the same foot length, substantial variation exists in overall shape, highlighting the limitations of traditional sizing systems based solely on length and width. This thesis demonstrates that statistical shape modeling and unsupervised learning can uncover underlying structure in 3D shape data and guide anatomicallyinformed fit options. The modeling framework developed here has direct applications in footwear design, enabling more precise, comfortable, and population-representative products.