<p>Distance-based outlier detection methods are widely recognized for their generality and effectiveness in identifying anomalies in multimodal data. As industrial datasets continue to grow in scale and complexity, such methods face increasing challenges in terms of computational cost and runtime efficiency. This paper presents STOD, an outlier detection algorithm that achieves substantially higher efficiency than iORCA by introducing a redesigned cutoff threshold update strategy and an improved pruning rule. STOD accelerates detection by reordering the outer and inner traversals over the dataset and data blocks, and by treating <i>t</i> data points as an independent block to update the cutoff threshold in advance. The updated threshold is then combined with the pruning rule to eliminate a large number of non-outlier points, thereby speeding up the detection process. In addition, STOD adopts a new starting point of the search order and an optimized search order, which further reduces the number of distance computations during detection. Experimental results on multiple real-world datasets show that STOD improves detection speed by 42.77% on average and reduces the number of distance computations by 32.38% compared with iORCA.</p>