We built a model that considered the entire inventory - both online and offline. With orders that were placed, orders that were ready for shipment cancelled orders, orders reserved at the store, and returns called out, the ‘available-to-sell’ inventory was published for every store manager’s perusal, and order planning. Order routing, thus, became more intelligent.
We then merged the online and in-store data sets to create one unified data lake that provides a rich source of insights for sales projection - without the additional effort of an operations team working to unify and prepare data for sales planning, week after week.
Several customers bought products online, especially during promotions, but returned some of it thereafter. The retailer relied on the number of products shipped to arrive at the demand metric. The proposed model helped understand demand more holistically. It offered better visibility into orders placed, orders cancelled, and orders shipped – thus leading to better demand forecasting.