Identification of right brokerage to the right distributor in mutual fund industry using machine learning techniques

Brokerage in finance industry plays a very important role for any organization. For a distributor, brokerage band is the crucial factor in determining their commission structure. Traditionally, the band was decided manually based on predefined criteria, which can be bias, subjective and inefficient. This research provides a data driven approach to determine brokerage band for distributor by using Machine learning algorithms. By using dataset containing Asset under management (AUM), geographical region and historical sales, we apply unsupervised algorithm to group distributors with similar traits into clusters. These clusters represent different brokerage bands based on distributor’s ability to contribute to an organization Sales. K-means, K-Prototypes, Gower distance with Agglomerative Clustering, and Gower distance with DBSCAN are used to create clusters based on feature similarity. The efficiency of these algorithms is measured using the Silhouette score and qualitative analysis of the resulting clusters. Using K-means, the Silhouette scores range from 0.67 to 0.73 across all branches, whereas K-Prototypes and Gower with Agglomerative yield scores of 0.88 and 0.75, respectively. These result shows that the clustering approach can segment distributor into meaningful clusters, which can be mapped to brokerage bands and help to optimize the commission strategy. These methods not only use for deciding brokerage band but also helps to gain deeper insights into distributors product preferences and the sales pattern.