Demand Forecasting: 3-Wk Pilot Implementation

Tredence Inc

Enable accurate demand forecast at a very granular level to enable increased transparency and efficiency in the supply chain planning and operations

Objective: Develop a ML based solution which can accurately forecast demand to optimize the planning of inventory, warehouse operations and logistics

Key Challenges Addressed:

  1. Internal factors are missed while planning for labor including tools and resources - no of plants / no of production lines / open order items.
  2. Lack of centralized consumption layer to visualize the current state and record the shortcomings.

How do we address your challenges: The long term forecast accuracy is improved using ML models by taking into account internal and external factors. A safety pool is kept to protect against variability. Open order data is taken into account while building the models. A frontend dashboard helps to visualize the AS IS state and TO BE state.

Pilot Outcome: Scope: ~10 SKUs, 1-2 distribution centres

2/4/8/16 week demand forecast generated by SKUs

Implementation Plan The break-up of the implementation plan is as below: Week 1 - Conduct data discovery and sanitize require data elements Week 2 - Develop ML models Week 3 - Fine tune forecasts and generate output

This implementation uses the following native Azure components: Azure Data Factory ADLS Gen 2 Azure SQL Database Azure ML Power BI