Organizational Clustering Using Fuzzy C-Means for Business Unit Performance Benchmarking

Business unit performance benchmarking is considered an important management control technique used in multinational corporations. The general methodology used here is the application of the clustering algorithm (K-Means and hierarchical agglomerative clustering). The application of this method leads to a rigid assignment of units to different performance classes and may eventually overlook an important category of business units that could fall into several tiers at once. The Fuzzy C-Means (FCM) clustering model proposes a solution to this problem. The method suggests that instead of binary assignment, each business unit gets a degree of membership μ_{ik}$\in[0, 1]$for each cluster k. The current paper demonstrates the implementation of the FCM model to the analysis of the performance of 2,840 business units from 18 multinational corporations (manufacturing - 8, finance - 6, technology - 4) over the period of 3 fiscal years (FY2022-FY2024). The research uses 28 KPIs. FCM with five clusters$(\mathbf{c}=\mathbf{5})$, fuzziness parameter$\mathbf{m}=\mathbf{2. 0}$, convergence threshold$\varepsilon=0.0001$, and a maximum of 300 iterations achieves an Xie-Beni (XB) validity index of 0.284-a 41.3% improvement over K-Means (XB: 0.484)-and a partition coefficient$\mathbf{P C}=\mathbf{0. 8 4 2}$. The five identified clusters range from High Performance (C1: 18.4% of units; mean performance index 84.2) to Underperforming (C5: 7.6%; index 28.4), with the portfolio weighted mean of$61.7 (\mathbf{0. 1 8 4} \times \mathbf{8 4. 2}+\mathbf{0. 2 4 8} \times \mathbf{7 2. 4}+\mathbf{0. 3 2 4} \times \mathbf{5 8. 4}+\mathbf{0. 1 6 8} \times \mathbf{4 2. 8}+\mathbf{0. 0 7 6} \times \mathbf{2 8}$.$4=61.7)$. A total of 612 business units (21.6% of the portfolio) exhibit maximum cluster membership below 0.70, identifying them as genuine performance boundary cases warranting individualised benchmarking analysis rather than tier-level intervention.

Hali tarjima qilinmagan