In today’s competitive retail landscape, accurate demand forecasting and inventory management is the cornerstone of success. At Target, data science drives this precision by leveraging AI and machine learning to optimize operations across its extensive retail network. With fully automated and integrated systems for forecasting, purchasing and product positioning, Target has significantly reduced manual interventions, thereby improving operational efficiency and ensuring seamless product availability for consumers, says Sharad Limaye, Senior Director, Data Sciences, Target. In this editorial interaction, Sharad shares details on how these systems were implemented, the collaborative strategies that drive their success, and the emerging innovations in AI and data science that are poised to redefine inventory management and demand forecasting for the retail industry.
Some edited excerpts from the interview:
How do AI and machine learning models improve the accuracy of demand forecasts?
Given the size and scale of our operations, AI and ML models are absolutely essential to improve the accuracy of demand forecasts. Target sells thousands of products in its stores and online through target.com. Generating accurate demand forecasts for millions of item-location combinations requires algorithms that are fast, scalable, explainable, and flexible. Target uses generalized additive mixed models (GAMM) to generate highly reliable demand forecasts across different sales channels and across the entire network every day for short-term execution as well as long-term planning.
ML models have helped significantly reduce forecasting errors and improve accuracy compared to our existing systems and processes, across all of our sales channels.
What is the role of data science in optimizing inventory management, from warehouse stock levels to shelf placement in stores, ensuring a smooth flow of products to meet consumer demand?
Data science is at the heart of decision-making when it comes to designing and developing solutions for all areas of retail, including inventory management. Some of the ways data science is used:
Data science models are fully integrated into decision-making to optimize inventory flow throughout the supply chain lifecycle. These models generate optimal policies that help decide how much inventory to purchase and place in warehouses and how much to replenish in stores. It’s very important to have the right balance here so you don’t buy too much (because storage space is limited) or too little inventory (because they might be out of stock).
Within stores, space and planning models enable better planning and positioning of items within stores taking into account store size, category affinities, expected sales and other operational constraints. Allocating optimal storage space to all categories is very important to boost Target’s sales.
Data science models help make smarter decisions upstream and downstream, ensuring smooth movement of inventory throughout the supply chain and into our stores, improving operational efficiency.
How has Target implemented fully automated systems for forecasting, purchasing and positioning, reducing manual interventions and improving operational efficiencies throughout the supply chain?
We built algorithmic solutions and data science models to power a fully automated and integrated forecasting, buying and positioning system that lays the foundation for modern inventory management at Target.
It all starts with understanding customer demand by considering several factors related to promotions, events, holidays, etc. By combining all of these factors, ML and deep learning models help us obtain an accurate estimate of unconstrained demand across our network of 2,000 stores. This powers replenishment and purchasing systems in which different algorithms seek to make the best possible decisions in the face of uncertainty. These are complex, multi-level inventory optimization problems where decisions must be made for millions of item-location combinations every day.
The integrations we have built into our production systems give us end-to-end visibility into various decisions and events and the flexibility to add more capabilities and enhancements as we modernize our supply chain. As this significantly improved the explainability of the models, it also helped reduce manual interventions over time, which was otherwise very difficult with our legacy legacy systems.
Can you talk a little bit about the need for cross-functional collaborative efforts to develop and scale these AI-based solutions?
Forecasting and inventory management systems are the bread and butter of any retail organization and require integration with complex systems that must operate with clockwork precision and are completely synchronized. To achieve this, data sciences work closely with our partner engineering, product and business teams to build and evolve these models. It is very essential to bring the right process knowledge and experience, as well as knowing what capabilities to develop with the right level of engineering and science. The teams work very closely and iteratively to build these systems. Building large-scale AI-driven production systems requires such joint planning and execution,
What are the upcoming innovations in AI and data science that could further improve inventory management and demand forecasting in the retail industry?
As supply chains become more complex than ever, the speed and accuracy of decision-making must also evolve. Simulation-based optimization, Markov models, bandit algorithms and reinforcement learning are all areas of research that will improve complex and sequential decision making on problems of a stochastic nature and algorithms facing a great variability.
Leveraging additional signals from operations to improve intelligence, making algorithms more explainable, faster experimentation coupled with suitable platforms that operate at high speed and scale will all be areas of hope for improved management inventory and demand forecasting.