A leading large-scale online grocery and food retailer , needed to improve inventory planning across its Distribution Centers (DCs) and Dark Stores (DSs). Existing forecasting approaches did not differentiate between DC and DS requirements, resulting in excess inventory and wastage on one hand and stock-outs and lost orders on the other. LTM re-engineered the forecasting solution using a customized Newsvendor model with Machine Learning, enabling independent demand forecasting at the DC and DS level. Built on Big Data, AWS Redshift, and Spark, the solution automated forecast transmission and scaled across 150+ locations, creating a foundation for PAN-India inventory optimization.
New Productivity Paradigms
We transformed inventory planning from a generalized, manual forecasting process into a location-specific, data-driven intelligence model—enabling the business to anticipate demand more accurately, reduce wastage, minimize stock-outs, and optimize inventory across its distribution network.
We owned outcomes by delivering:
Reduced excess inventory and associated wastage through more accurate demand forecasting
Reduced out-of-stock occurrences, improving product availability and customer experience
Automated transmission of forecast data and performance monitoring
Scalable implementation across 500 SKUs per location and 150+ locations
Foundation framework designed to extend to approximately 25,000 non-perishable SKUs
Our Client
The client is a leading large-scale online grocery and food retailer operating across multiple urban markets in India. The organization manages high-velocity consumer demand and complex supply chain operations, with a strong focus on product availability, freshness, and fulfillment efficiency.
With a large network of Distribution Centers and Dark Stores, effective inventory planning is critical to balancing product availability against storage costs, wastage, and customer demand.
Challenges
The existing inventory planning approach did not sufficiently differentiate demand requirements at the Distribution Center and Dark Store levels, creating inefficiencies across the supply chain.
The existing inventory planning approach lacked location-level demand visibility, making it difficult to differentiate requirements across individual Distribution Centers (DCs) and Dark Stores (DSs).
Limited ability to anticipate SKU-level demand and potential stock-outs was leading to inconsistent inventory availability across locations.
Overstocking was driving excess inventory and wastage, increasing storage costs and tying up working capital.
Stock-outs were resulting in lost orders and missed sales, impacting fulfillment performance and customer availability.
Demand forecasting at the DC and DS level was largely manual and lacked a robust analytical foundation, limiting planning accuracy.
Manual processes for transmitting and forecasting inventory requirements were time-consuming and prone to inefficiencies, slowing replenishment decisions.
Forecasting was constrained to a limited set of SKUs and locations, making it difficult to address the complexity of the broader network.
The existing approach was not designed to scale, , limiting the ability to consistently balance inventory availability, storage costs, and demand variability across locations.
LTM's Solution
LTM re-engineered the client's forecasting solution using a customized Newsvendor model enhanced with Machine Learning to enable more accurate and location-specific inventory planning. Key aspects of the solution included:
DC and Dark Store-Level Forecasting
The solution independently forecast inventory requirements at both Distribution Center and Dark Store levels, enabling more precise planning based on the unique demand characteristics of each location.
This helped the business better anticipate:
Potential stock-outs
Excess inventory
SKU-level demand variations
Location-specific inventory requirements
Machine Learning-Enhanced Newsvendor Model
LTM customized the Newsvendor forecasting approach using Machine Learning to improve demand estimation and help balance the trade-off between product availability and inventory wastage.
The approach enabled the organization to make more informed stocking decisions based on expected demand.
Big Data and Cloud-Scale Architecture
LTM redesigned the forecasting solution using:
AWS Redshift
Apache Spark
Big Data technologies
Mindtree's Big Data ETL framework
ABC (Audit Balance Control) framework for data integration and automation
This provided a scalable architecture capable of supporting high-volume forecasting across a growing network.
Automated Forecast Distribution
An API framework was implemented to make forecasts available to downstream applications, eliminating manual transmission of forecast data and enabling automated consumption across supply chain processes.
The solution also enabled performance monitoring at the DC and DS level to improve visibility into forecasting effectiveness.
Business Benefits
The intelligent forecasting solution delivered measurable improvements across inventory planning, operational efficiency, and customer experience:
Enabled independent inventory planning at the DC and DS level, improving supply chain efficiency and decision-making.
Reduced inventory wastage and associated costs by minimizing excess stock and dump scenarios.
Improved product availability through more accurate demand forecasts and reduced out-of-stock occurrences.
Reduced order losses and improved customer satisfaction through better SKU availability.
Eliminated manual forecast transmission through automated data integration and downstream APIs.
Improved visibility into forecasting performance and effectiveness at individual DC and DS locations.
Scaled the solution across 500 SKUs per location and 150+ locations as part of PAN-India implementation.
Established a reusable foundation capable of extending forecasting to approximately 25,000 non-perishable SKUs.
Created a scalable data and analytics framework to support continued supply chain transformation.
Conclusion
By combining Machine Learning with a customized Newsvendor forecasting model and a scalable Big Data architecture, LTM helped the client transform inventory planning into a more granular and intelligent supply chain capability. The solution enabled independent forecasting across Distribution Centers and Dark Stores, reducing the trade-off between excess inventory and stock-outs while automating critical planning processes. With a scalable foundation for PAN-India deployment and expansion across thousands of SKUs, the client is better positioned to optimize inventory, improve availability, and deliver more efficient fulfillment at scale.
Ready to power your inventory planning with AI-powered demand forecasting? Contact us at info@ltm.com