scroll

How we helped a leading manufacturer of paints reduce average monthly sales loss by 22% and reduce product wastage by 14% through a scientific forecasting approach

Challenge

INDUSTRY OVERVIEW

A leading paint company in India, has a well established and efficient supply chain that supports the company's operations and helps it meet customer demands efficiently. As a crucial aspect of its supply chain management, demand forecasting plays a vital role in ensuring the smooth delivery of raw materials and finished products. The existing system uses historical data from SAP analytics system on which business logic is applied to determine the forecast.

It intends to leverage the power of machine learning (ML) and automate its forecasting process to consistently improve the accuracy of demand forecasting, reducing under forecasting, and the subsequent loss of sales percentage and inventory levels.

CHALLENGE

To develop an automated forecasting system at the starting of any month and a forecast correction process every 5 days in a month to detect any spike of sales and enables the company to adjust their forecasts, accordingly.

Gaps in the current system:

  • The current method of forecasting is defined by business logic and varies from SKU to SKU. The process in manual and is prone to errors / misses.
  • SKUs have separate rules for separate categories for the adjustment of the forecast every 5 days based on ending promotions and discounts.
  • The existing system is very dynamic in nature with lot of manual intervention to alter the monthly forecasts looking at the sales trend for the month and other external factors.

Why were we brought in?

Ganit hasmade a significant dent in various industries using data science and analysis. Ganit partners with clients to translate their data into a tangible, insightful plan of action that delivers on a measurable impact to the clients’ topline & bottom-line growth.

Our approach

Methodology

Required data from the SAP system is migrated to S3 on a daily and monthly basis.

Data Processing is performed on the data received and TTS, RTS and meta data files are generated for forecasting in Amazon Forecast. 3 months forecast is being generated at the beginning of every month which are formatted in required formats before storing the final forecast file in S3.

Every 5 days, MTD forecast to MTD sales ratio is calculated. Factor based on this ratio is used to adjust the forecasted numbers. Summary files are generated at different level for the business to review and take important decisions along with the final correction file which are stored in S3.
An automated pipeline to deploy AWS glue jobs & Amazon Forecast are developed in combination with Eventbridge and Lambda functions.

Features of the tool

  • Automated notifications over email are configured on success or failure of the process and on generation of deliverable files.
  • Automation has allowed the client to focus on multiple critical tasks rather than just one.

A valuable difference

Impact

  • Accuracy of the forecast increased from 60% to 85% and further to 95% with the 5-day correction process.with the 5-day correction process.
  • Average monthly loss of sales reduced by 22% (approximately INR 2.7 crore).
  • Over-forecasting has been reduced by 12.5% (value basis) and 14% (volume basis).
  • 6 hours reduction in time taken for forecasting of more than 60,000 units.
Success stories

See the impact that we make on our
cross-industry client base.

Top