top of page

Demand Forecasting for the retail

arm of a Fortune 100 manufacturer

Jun 2018, Sydney

Eamonn Barrett

Increase stock availability through accurate demand planning with a long supply chain


Short product life cycles, long supply chains.


AutoML Demand Forecasting methods

Our client was the retail arm of a multinational manufacturing client who wished to bring advanced demand forecasting to their operations. A lengthy supply chain combined with short product life cycles made for a complex dynamic, requiring long term forecasts that went far beyond the horizons of most retailers.


Remi AI was engaged to build a forecast pipeline for sales 9 months in advance. The complexities of such a forecast horizon will make most data scientists either go pale or lick their lips in delight (false dichotomy? maybe), and the team at Remi leapt at the opportunity. From the outset, there were two elements that were key to the success of this project:

  • Seasonality: Our models had to pick up the change points in demand within a given time period

  • Product Matching: Our models needed to be able to handle replacement products which was done via clustering


The team at Remi AI was able to produce some exciting results on training data with accuracies averaging 25% MAPE across the product range. Given the long forecast horizon, the results will be verified as we move through the 9 month period - early results are looking excellent!

bottom of page