Effective agent selection for insurance provider


South East Asia-based insurance provider catering to small and medium businesses


The sales and marketing groups were facing four critical business issues: High cost of acquiring and training agents, high attrition rates, inconsistent performance of agents, and time-consuming agent selection process.


SG Analytics deployed a four-step process to address this business problem:

  • The SG Analytics team first built an agent database using parameters such as written premium, tenure of service, conversion rates for high value customers and persistence index
  • Based on the data, the SG Analytics team created three clusters of agent performance using the centroid-based method and hierarchical clustering at both portfolio and regional level
  • The SG Analytics team then used statistical techniques like ANOVA, chi-squared test of independence and various other descriptive statistical techniques to identify characteristics for reporting
  • Finally, the team developed a model using supervised learning techniques such as Random Forest, SVM, NeuralNets and Discriminant Analysis


85% accuracy on unseen data.
Improved the effectiveness of the agent selection process.
Increased the overall hit ratio by 17% and revenue by 12% on a sequential basis.