Supporting underserved communities with better access to grocery and consumer goods
ElasticRun helps small businesses improve their supply chain for grocery and consumer goods in rural areas
Machine learning models managed
Fewer data pipeline slowdowns from large data volumes
In cost savings on more than 20 systems
ElasticRun, India’s only B2B e-commerce platform, connects brands with millions in rural family-owned businesses. The brand aims to target these hard-to-penetrate markets and drive growth in underserved areas of India. ElasticRun needed a scalable, fail-safe and reliable data and processing infrastructure to help optimize their supply chains, speed delivery of products and improve store-level demand forecasting. ElasticRun implemented Databricks for scalable data storage, real-time analytics and machine learning — leading to a 90% reduction in data pipeline slowdowns and cutting IT costs by 33% across their data processing infrastructure.
Hindering economic upliftment due to data bottlenecks
Focused on expanding access to India’s rural consumption market, ElasticRun serves “kirana” stores, or small, family-owned businesses. By providing access to a wider range of affordable and diverse products, these businesses offer an essential lifeline to many low-income families in rural areas of India. Beyond a certain scale, the e-commerce platform’s open source data infrastructure, ElasticRun Connect, required multiple scale-ups and refactoring. This led to pipeline delays that occurred at least twice a month due to the sheer volume of data, from store profiles to transactions and inventory counts. As ElasticRun’s business grew, the data inundation put additional pressure on the data pipelines and on the consumption of timely insights, which put additional pressure on OLTP systems.
Another critical area for improvement was ElasticRun’s supply chain software, ElasticRun Logistics, which required real-time data to help customers make smarter decisions around inventory management, shipment tracking and demand forecasting. The existing data infrastructure exhibited limitations that needed to be unlocked to handle the increased data load from ElasticRun’s growing customer base. “Our existing pipelines and data infrastructure had hit a glass ceiling. We needed a reliable data and AI platform that could meet our current data demands and scale as we continued to grow,” Yogesh Kulkarni, Senior Director of Technology at ElasticRun, explained. His team recognized that without a stronger, more scalable platform, they risked falling behind in delivering better customer outcomes.
Additionally, ElasticRun’s team managed data from multiple systems (e.g., ERP platforms and logistics software) and required a solution capable of supporting both batch and streaming data ingestion, managing thousands of machine learning models and enforcing data access policies. A future-proof, scalable infrastructure was essential to support the cost-sensitive needs of India’s rural market.
Centralizing data to improve customer experiences at scale
ElasticRun’s rapid growth demanded a data and AI solution that could keep pace with their expanding needs. Adopting the Databricks Data Intelligence Platform provided the ideal solution, with Delta Lake at the core of their suite of products. Delta Lake’s flexible medallion architecture, a data design pattern that organizes data into progressively refined layers — Bronze, Silver and Gold — enabled ElasticRun to process both real-time and historical data while eliminating the pipeline failures that previously hindered operations. Databricks’ lakehouse architecture also allowed the company to store live data for instant analytics while maintaining low-cost archival data for future access. As Kulkarni explained, “Delta Lake gave us the ability to scale effortlessly, allowing us to focus on driving real-time insights instead of worrying about system failures.”
To further streamline their data processes, ElasticRun deployed Databricks Delta Live Tables to automate data transformations across various environments — development, staging and production. This automation removed the need for manual intervention, which had previously been a bottleneck, and ensured data accuracy and timeliness. Feeding this clean, reliable data directly into the lakehouse supported ElasticRun’s demand for real-time insights, particularly in demand forecasting and inventory management to prepare for seasonal peaks and other trends.
With support from Databricks, the ElasticRun team developed a strategy that identified and prioritized jobs for migration, established workflows and implemented necessary downstream application changes. This collaborative effort ensured a seamless transition to the Databricks Data Intelligence Platform, enabling the team to successfully migrate 80% of their jobs within just six months.
Additionally, Databricks Workflows streamlined data ingestion, transformation and analytics to help ElasticRun’s data team increase pipeline reliability. With the assistance of Workflows, the company was able to adjust quickly to shifts in customer demand or supply chain changes. For instance, Workflows supported real-time adjustments to inventory levels based on sales data to guarantee optimal stock without overstocking. It also helped better orchestrate delivery routes in the company’s logistics network and, consequently, reduced transportation costs and improved delivery times in the rural markets ElasticRun’s clients serve.
Machine learning played a pivotal role in ElasticRun’s operations, with over 10,000 models in production, focused on optimizing supply chains, personalizing and improving demand forecasting. MLflow, designed to manage the complete machine learning lifecycle, enabled ElasticRun to track experiments, package code, deploy models across environments and share them efficiently across teams to continuously enhance these processes. Plus, Databricks SQL and Unity Catalog provided secure data access and governance by allowing over 1,500 internal users to query data while maintaining strict compliance with consumer data policies. All in all, the seamless integration of machine learning and analytics positioned ElasticRun for sustained growth in India’s rural consumption market.
Decreasing pipeline failures while increasing NPS
Adoption of the Databricks Data Intelligence Platform transformed ElasticRun’s data operations, resulting in a 90% reduction in pipeline slowdowns and infrastructure incidents. By leveraging Delta Lake alongside Delta Live Tables and Databricks Workflows, ElasticRun increased the reliability and availability of their data pipelines. The boost in performance allowed their internal teams to access real-time data without interruption, which directly led to a 25% improvement in their internal net promoter score (NPS).
With a stable data infrastructure in place, ElasticRun helped optimize supply chains and personalize offers for kirana stores more effectively. MLflow, managing over 10,000 machine learning models, paved the way for more accurate demand forecasting and anomaly detection to remedy operations and improve customer satisfaction. “By consolidating 2,500 reports into 50–60 dashboards, we’ve shifted our focus from troubleshooting to innovation,” Kulkarni said.
Looking ahead, ElasticRun is set to expand their use of predictive analytics and machine learning, further improving internal decision-making and customer experience. One of their future key initiatives will involve exploring the potential of generative AI models, particularly language models that analyze customer and vendor interactions in Indian languages. In their operations, Databricks will remain a key partner, driving innovation and supporting their mission to tap into India’s $600 billion rural consumption market.
