One of the most complex and diverse products in the buyers’ market is the garment item. This is due to the multifaceted problems with fitting leading to poor sales performance in retail and increased garment return rate. This study proposes an algorithm based on customer pair wise matrix for improving the fit of the garment size system. The proposed algorithm is applied successfully to an anthropometric dataset consisting of 286 female Corp members. Its performance was compared with the existing KMedoid algorithms using the aggregate loss of the fit function and a novel percentage degree of fit function. Analysis of the observed result using t test statistics suggests a statistically significant difference at (t(18)=5.728, p=2.46E-05) and (t(18)=5.188, p= 7.4E-05) exist between the percentage degree for the proposed algorithm and the K-Medoid algorithm by Spath and Kaufman and Rousseeuw respectively. A similar result was obtained for the aggregate loss of fit. The algorithm enables a balance between percentage degree of fit and number of size groups for a target population
