Demand Forecasting

Smart demand forecasting reduces operational costs for CPG company

» Smart demand forecasting reduces operational costs for CPG company

The client is a British multinational health, hygiene and nutrition company delivering across different product lines to 200+ geographic markets. They manage a complex supply chain with 40,000+ retailers across the globe.

Business Need

The CPG company manages a complex supply chain with 40,000+ retailers. They need accurate demand forecasting to maintain SKU availability across regions. To predict weekly demand, procure raw materials and schedule production, they used multiple data pools in SAP and JDE systems.

Factoring in parameters like sales, trends and seasonality, they generated weekly reports for forecast quantity vs. actual sales. However, these manual reports turned out to be inaccurate and incomplete. Due to a lack of unified visibility into system metrics including logistics and stock level, they faced internal conflicts across functions.

Hence, they partnered with Hitech Analytics team build a robust demand forecasting model to:

Improve forecast accuracy
Timely respond to critical concerns of inventory optimization
Procure right amount of raw materials
Gain real-time stock visibility

A thorough root-case analysis by Hitech Analytics data scientists of their existing forecast model revealed challenges like:

Lack of visibility of historical data across multiple product lines to make accurate forecasts
Absence of mapping between the data across marketing insights, customer databases, supply systems and product lines.
Inefficient information system to meet varying customer demand trends in different markets like US and UK.
Lack of information of seasonal parameters, volume sold on deals (VSOD) and depth of discount (DOD) led to inflexibility to make dynamic changes to the existing data models.
Erroneous inventory data, inefficient analytical tools and inability to understand data to maintain product standards.

Want to extract value-based insights from large amounts of data generated by your supply chain?

Get in Touch with us »

Solution

Hitech Analytics team delivered a customized demand forecasting solution to the client that helped to:

Accurately map sales, brand and promotional data
Seamlessly integrate all supply chain operations like product availability, management and operations for suppliers to end-customers.
Enhance their existing data systems through data pre-processing activities like cleansing and standardization.

Approach

We assessed their existing systems and conducted in-depth discussions with various brand managers to gather data, its sources and actual business problems.
Gathered legacy demand data lying in siloed sources for unified understanding of impact rising due to trends, seasonal, cyclical and random components.
Experimented with various time series forecasting algorithms and zeroed down on the best options using ranking approach from measure of error parameter like MAD, MAPE and RMSE.
Identified gray-areas in existing data and deployed course-correction for it. For example: added cross data validation and feature selection for value-based forecasts
Paced up existing historical datasets using data pre-processing techniques and applied classical time series algorithms for better forecast accuracy.
Deployed automated smart forecasting support to interact with existing numerous applications and maintain accuracy in store-level forecasts.
Synchronized the code to scale up for different countries with minor changes.
Our model was successfully tested in USA region for six months and further replicated in different markets like Australia, Brazil, France, Mexico and UK.

Results

Increased real-time stock visibility in a regular and volatile environment
Better forecasting accuracy
Improved inventory management for all products across stores
Enhanced supply chain efficiency