How clustering helped a jewellery major predict sales at a store level

Inventories in store comprise the largest investment in a fine jewellery business. For a chain of jewellery stores, the issue was to identify what SKUs to stock, in what quantities and in which outlet. The objective was to optimize the inventory at a store level. The secondary objective was to improve the new product indents for these stores given the history of sales by product types.

The approach adopted was to cluster stores based on the following variables – Items sold, type of items, days taken to sell and quantity sold. Using this clustering approach the stores were divided into four types, for which inventory rules were created for replenishment. This led to lower inventory levels at the store level, while improving the sales as the number of variants at the store increased and there was a better sync with sales.

‍

Studying the clusters at a product group level led to insights on the likely sales of new products. This also helped in the distribution of new products, as opposed to the earlier method of trying all stores and course correcting over time. 

‍

The benefits from this exercise – lower inventories, improved product mix, increased sales, fewer discounts, lesser stock-outs. The other major benefit was that these clusters were then taken and profiled at a customer level from primary research and store loyalty programs enabling a better understanding of the customer behaviour at a particular store type. This input was invaluable in creating better customer promotions, new merchandise and in picking merchandise mix for new stores in similar areas.

‍