Locate, group and up goes performance

This study was based on work done for a bank with rural focus. Most of the branches were established in the southern state of Tamil Nadu, India.

Banks are continuously opening branches in villages. Several (4-10) villages were grouped together to form a cluster. Every cluster had a manager + team to execute day-to-day activities. The cluster manager typically needs to travel between the villages belonging to his/her cluster.

Motivation  Performance (Average Cluster Revenues) is directly related to the ‘tightness of the cluster’. If the villages in a cluster are nearby or tight, then the average cluster performance is higher.

Average Cluster Revenue vs Distance within Cluster

Note: Fictitious (but representative) data presented here.

READ MORE : Meri Maggi Story – May 2015 | Power cuts in Bangalore Sept 2015

Question   If the performance of the company is closely related to cluster tightness; group villages in a way improve performance

Data   Collected from ~150 centers belonging to 60 clusters (which evolved as the company grew).

Predictive Analytics Method   Dendrogram Hierarchical clustering.

Initially, each village is assigned to its own cluster and then the algorithm proceeds iteratively, at each stage joining the two most similar villages, continuing until there is just a single cluster. At each stage distances between clusters are recomputed.

Results Summary  
Note: Fictitious (but representative) data presented here.
After Clustering
A simple method like clustering can add enormous value in improving and predicting performance.

Leave a Reply

Your email address will not be published. Required fields are marked *