K-means clustering is a fundamental technique in unsupervised learning, renowned for its straightforwardness and effectiveness in organizing datasets into clusters based on feature similarities. Despite its widespread acceptance, the algorithm faces a significant hurdle: its sensitivity to the initial placement of centroids, which often leads to inconsistent clustering results. This study delves into a comprehensive analysis to assess how different methods of initializing centroids impact the performance of K-means clustering. Through meticulous examination of inter-cluster and intra-cluster distances across a diverse array of datasets, this research provides a detailed quantitative evaluation of clustering efficacy under varying initialization conditions. The results of this analysis shed light on a critical aspect of K-means clustering: its performance, as gauged by accuracy and correlation metrics, is notably affected by the positioning of initial centroids. This observed variability underscores the necessity of strategic centroid initialization in enhancing the reliability and effectiveness of the algorithm.