A Multi-chain Food & Beverage and a Fast-moving Consumer Goods firm
Project Title
On-premise Data Lakehouse along with a Sales forecasting model
Client
On-premises Data Lakehouse with a Sales forecasting model for a Consumer Food & Beverage Firm
Problem
Long Processing Time: Aggregation queries across multiple SQL views can take 3–6 hours to execute.
Operational Bottlenecks: Decision-making is delayed for procurement, kitchen planning, and staffing.
System Load: SQL servers are heavily strained, affecting other transactional systems.
No On-Demand Forecasting: Forecasts cannot be generated dynamically or exposed in real time to business users.
Lack of Automation: The existing process does not support rolling forecasts or model-based projections.
Solution
We deployed HDFS for scalable storage and Apache Hudi for managing transactional, versioned datasets with support for incremental updates and time-travel queries.
We built a custom AI-based sales forecasting model using ARIMA, Prophet, XGBoost, LSTM.
Forecast outputs and curated datasets were made available to business users through BI tools (Tableau & Qlik), powered by Dremio’s virtual datasets.
For fast and interactive analytics, we integrated Dremio as the query engine, enabling business users to explore massive datasets using familiar SQL without relying on exports or intermediaries.
Value Proposition
Drive SKU-level demand forecasting at each outlet with 95% accuracy
Improve branch-wise inventory control by 30% less
Support production and procurement planning
Minimize wastage and ensure availability of fast-moving items
