This paper seeks to evaluate whether the chemical profiling data obtained with a constant weight approach can be used in the classification of highly cut heroin samples. A constant weight of a seized substance instead of the conventional weight equivalent to 15 mg heroin base was used to profile the manufacturing impurities. The study attempts to optimize four clustering tools using 12 impurity peaks extracted from the heroin samples (52.3% purity) analyzed at 650 mg sample weight to find the most ideal statistical techniques for sample classification. The effectiveness of four clustering tools, namely the principal component analysis (PCA), hierarchical cluster analysis (HCA), K-means clustering (KMC) and discriminant analysis (DA) was assessed using 25 heroin samples derived from five known batches. HCA and DA proved promising in clustering the related samples. Finally, only the HCA was then employed to evaluate the general relationships between 46 unknown heroin samples.