Optimal store clustering for allocation to salesforce

Abstract

Everyday millions of products flow from factories to consumers via chain of warehouses, distributers and retailers. All players are critical in maintaining the supply and services to customers. With huge competition and thin margins, optimized decisions are mandatory in order to preserve margins. This project focuses on distributers and specifically salesperson in the supply chain. To optimize on fixed costs, generally, a salesperson partners with multiple retailers. He/she sells many items from same and different brands. Large number of SKUs with different size, demand & consumption distribution introduces complexity in the model. Every day salesperson must decide SKUs, quantity, retail outlets to deliver, order of delivery. There are many retail shops across a city, and they belong to different categories like Kirana shops, Supermarkets, Pharmacy, and so on. These shops have varied demands and must be visited frequently by the salesperson to match the supply with their demand. Usually, a dedicated salesperson is allotted to each retail shop and he/she takes care of the shop’s requirements. In this way, multiple retail shops are allotted to a salesperson and is made responsible for their order fulfillment. It is imperative here that the salesperson meet requirements of all the retail shops accurately and ensure timely and efficient delivery of products. This can only happen when there’s an efficient distribution of retail shops to each salesperson based on the location, category, demand, and frequency requirements of a retail shop and also the workhours available with respective salesperson. This efficient distribution of retail shops to salesperson can be achieved by clustering all the retail shops into different categories based on the criteria discussed above. Given all retailers, who are customers of any brand, the objective of the project is to cluster the retailers into the most optimal groups such that a single group can be served by a single salesperson. In doing so, it is crucial to ensure that the size of all clusters is approximately equal and there is an equivalent allocation of different categories of retail shops in each cluster. For example: the number of retail shops in any cluster can vary from 50-60. However, it needs to be within that range. Also, a cluster should not be dominated by a particular category of retail shops and not all supermarkets/Kirana shops can be in the same cluster. The model must adhere to various constraints faced by a salesperson. All these constraints together make the clustering more challenging and results in a need for algorithms other than the traditional clustering methods.