A leading Frozen Food Manufacturer saves $20M annually due to Accurate Demand Forecasting

About

The client, a leading North American frozen food company manufactures several frozen food products and potato specialties across the world.

Challenges

  • The manufacturer faced challenges to accurately predict the demand using traditional / conventional methods on an SKU level for 12000+ stores across USA. This forecasting method did not yield accurate results in a dynamic environment.
  • The client was losing a minimum of 6% of their revenue due to forecasting inefficiencies at the retail level. It translated to close to $50 Mn dollars.

Solution

Tenzai developed a custom ML based forecasting solution for the client by implementing our Purpose-Driven AI approach.

Internal data from enterprise data systems as well external data were utilized for the developing the model.

External data considered included seasonality trends, holidays, retailer and store data, weather and macroeconomic indicators.

The analytical DataMart was built on AWS Redshift and the machine learning environment was created on AWS EC2.

The forecasting model was an ensemble of multiple algorithms like Random Forest, LSTM and FB Kats. The model performance was measured using metrics like MAPE and model accuracy was tracked across various granularities.

The forecasting model was made accessible to the business through an intuitive dashboard. It helped business users to analyze the forecasts across multiple regions, stores, categories and SKUs and make informed decisions.

Why not use data to gain insights, guide efficiency efforts, and make predictions? Contact us.

Results

Post-implementation, the client experienced a very high increase in terms of on-time collections across all regions.

The solution helped collection agents increase productivity, reduced collection expenses.

The solution helped the finance team to have better financial stability with predictable cash flows.
The other benefits include:

  • Forecasting accuracy improved by ~50% compared to conventional methods.
  • $20 Mn in annual savings due to efficient forecasting.
  • More than 15% reduced average inventory holding costs on a weekly basis.