Smart inventory management to build a smart retail store

Overview

Our client was a bespoke apparel designer store who were in the midst of expanding their retail boutique store in addition to their online operations. They managed multiple warehouses which handled storage and dispatch of both unstitched fabric and finished apparels. The business heavily relied on manual entry of inventory across departments and were bogged down by its sheer enormity. Coordinating the stock between the departments of unstitched fabric materials and designed ready-made was very daunting. They often found themselves running into situations of understocking or overstocking leading to inefficient production cycle. The client wanted us to design a AI-powered intelligent inventory system that can help optimize inventory at all levels in the demand chain. They were also interested the to analyze gap between the inventory requirements of instore and online store. They wanted to predict future buying behavior to detect and act on supply chain anomalies in a timely fashion.

Challenges

Data inventory management becomes a tedious process due to how much of it is usually accumulated by any company.The inventory managed here were at two different levels, design process and mode of operation. In the design aspect, it had both raw material and finished products and in the mode of operation, it was offline retail and online store. It is therefore, a challenge to track every item in the inventory and to get relevant analysis out of it. Keeping track and analyzing data from the offline retail store was tedious as most of the records were managed manually across departments.

Solution

The data requirements for the solution was quite straight forward as the business had already gathered all the sales records in their relation data bases. We defined a time series based prediction model that can understand the patterns from the previous sales data. The prediction model was able to reproduce the results with over 92% accuracy on the historical control data. This led us to trust the model and deploy on product, where the model was able predict on a simillar level of accuracy.

Impact

The inventory management model that we built was able to leaverage on the historical data to analyze supply and demand patterns. The prediction model was able to help them understand both in-store and online demand signals and plan their inventory accordingly. The predicted consumer demand was 85% accurate and they were able to deflect their inventory to appropriate locations with very high placement accuracy. The optimized smart warehouse operation brought in an increase in product visibility and 15% reduction in their operation costs. Walmart has recently introduced the Retail Intelligence Lab, an AI-powered store that allows companies to better manage their inventory in the store. https://softengi.com/blog/ai-in-retail-waiting-is-not-a-winning-option/

Technology stack