CPGi: Actionable Intelligence Platform for CPG Companies

Github: https://github.com/glombardo/CPGiV2

Problem Scope

The Consumer Packaged Goods industry operates on razor-thin margins where a 1% improvement in pricing or inventory management can translate to millions in profit. Yet most CPG companies still rely on disparate tools and Excel-based models for critical decisions. This project addresses that gap with a unified analytics platform that brings together demand forecasting, price optimization, inventory management, and marketing mix modeling in a single, intuitive interface.

CPG companies face unique analytical challenges. Unlike software or services, physical products require careful inventory management across complex distribution networks. Pricing decisions must account for competitor actions, promotional calendars, and price elasticity that varies by channel and region. Marketing spend needs optimization across traditional and digital channels, with proper attribution for delayed effects.

Technical Architecture

The platform leverages Python's data science ecosystem with Streamlit as the frontend framework. This choice enables rapid prototyping while maintaining production-grade performance. The modular architecture separates forecasting models, optimization engines, and visualization components, making it easy to swap algorithms or add new capabilities. Each module uses ensemble methods combining multiple approaches. For example, demand forecasting blends Prophet's seasonality handling with Random Forest's ability to capture non-linear patterns and ARIMA's time series fundamentals.

Key Innovations

What sets this platform apart is its focus on CPG-specific requirements. The inventory optimization module goes beyond simple EOQ calculations to handle multi-echelon networks with service level constraints. The pricing engine models cross-product elasticities and competitive responses, critical for category management. The Marketing Mix Model supports multiple methodologies (Frequentist, Bayesian, and Lightweight MMM) with channel-specific adstock and saturation curves, reflecting how different media types decay and saturate differently.

Upcoming Improvements

The platform's modular design enables continuous improvement. Planned enhancements include integration with major ERP systems, support for probabilistic forecasting to better handle uncertainty, and reinforcement learning for dynamic pricing. The open-source nature invites collaboration from the data science community to tackle industry-wide challenges like attribution modeling and demand sensing. As CPG companies face increasing pressure from e-commerce disruption and changing consumer behaviors, tools like this become essential for maintaining competitiveness in a rapidly evolving landscape.