Other links:

Other links:

Recent Advances in Data Science can effectively address issues faced by brick-mortar retailers

This research conducted by Prof. Anirban Mondal and his team addresses the issue of itemset (products bought together by consumers) placement in retail stores in a diversified manner so as to maximize retailer revenue

Imagine for a moment a world without retail stores. As a consumer, you would need to travel far and wide to obtain access to your desired products from the factories of multiple manufacturers, thereby expending both time and effort. 

Now, observe how retail stores provide a valuable service to society by making a wide variety of products from different manufacturers readily accessible to consumers under the same roof. This allows consumers the flexibility of choosing their desired products from a range of offerings from different manufacturers at various price points, thereby creating the convenience of one-stop shopping

Interestingly, even though online shopping has been becoming increasingly popular, brick-and-mortar retailers still retain their importance because consumers often prefer to have a shopping experience and get a “feel” of the product.

Brick-and-mortar retailers, however, face several pain points such as (a) reduced consumer foot traffic due to intense competition from online retailers, (b) disruptions in the global supply chain due to events e.g., natural disasters and wars, which adversely impact retail business sustainability (c) ever-increasing consumer expectations regarding shopping experience, (d) enormous amounts of complex data (both structured and unstructured) about products, suppliers, consumer purchase transactions and consumer data in the social media, and (e) lack of adequate retail analytics for converting the data into actionable insights. Notably, these challenges will continue to grow. 

Here, recent advances in data science can provide opportunities for effectively addressing these issues and improving retailer revenue. 

Anirban Mondal, Associate Professor of Computer Sciences at Ashoka University, has explored how retailers can diversify their product offerings to facilitate sustainable long-term revenue earnings. Diversification enhances business resilience in the face of a wide gamut of external macro-environmental factors (e.g., economic, legal, political, social, and technological) which could potentially impact the business environment. 

For example, the ongoing war in Ukraine has caused significant supply chain disruptions globally. A high dependency on the sale of a single product may cause the retailer to suffer a significant loss in revenue in scenarios, where a sudden drop in either supply or demand for the product occurs due to sudden macro-environmental changes. 

In medium-to-large shopping malls, consumers tend to buy multiple products instead of buying only a single product. In data mining, sets of products, which are frequently bought together by consumers, are referred to as itemsets e.g., {bread, butter, jam}. 

Observe how placing itemsets in the slots of retail store shelves enables the retailer to exploit the association among the items being sold and also helps the consumers to locate their desired products in one place. This can significantly improve retailer revenue. 

In essence, strategic placement of diversified and high-revenue itemsets in a retail store is critical to improving retailer revenue. 

However, the problem of generating such itemsets is essentially combinatorial in nature, hence there is an explosion in the number of candidate itemsets; which causes memory and processing time issues.

In this regard, Prof Anirban Mondal and his research group have addressed the issue of itemset placement in retail stores in a diversified manner so as to maximize retailer revenue; the work is published in Applied Intelligence. Researchers from Ashoka University proposed a framework for the retrieval and placement of diversified and high-revenue itemsets of various sizes on the shelves of a retail store.

‘Our extensive performance study with both real and synthetic datasets demonstrates the effectiveness of our proposed High-Utility and Diversified Itemset Placement (HUDIP) scheme in efficiently identifying and placing high-revenue and diversified itemsets, thereby leading to improved retailer revenue’, Prof Mondal shared. 

He further added, ‘In the near future, we plan to investigate the cost-effective integration of our proposed framework into existing retail IT infrastructure. We are planning to build a Cloud-based software tool to facilitate retail stores at scale towards placing itemsets based on the knowledge of consumer purchasing patterns extracted from historical purchase transactions.’ 

Strategy/Approach developed:

Our proposed kUI (k Utility Itemset) index efficiently retrieves diversified top-λ high-revenue itemsets. The kUI index comprises multiple levels, where the kth level corresponds to the top-revenue itemsets of size k. Instead of identifying and storing all possible itemsets, the kUI index averts the combinatorial explosion of candidate itemsets by limiting the number of itemsets stored at each level. Additionally, we propose the HUDIP (High-Utility and Diversified Itemset Placement) scheme, which exploits our proposed kUI index for placing high-revenue and diversified itemsets. 

(Edited by Dr Yukti Arora)

Reference Article: A framework for itemset placement with diversification for retail businesses, Applied Intelligence, September 2022 | Volume 52 | Pages 14541–14559

Authors: Anirban Mondal, Raghav Mittal, Parul Chaudhary & Polepalli Krishna Reddy

Study at Ashoka

Study at Ashoka