top of page

Demand Forecasting in Ecommerce

Feb 2019, Sydney

Eamonn Barrett

Long lead times and an international pandemic no longer need to mean chaos for your demand forecasting.

Remi AI was engaged to connect the client’s data into the forecasting engine in the Remi AI platform, with the  goal of forecasting sales up to 12 months in advance. Starting with a small sample of SKUs to prove the technology, the Remi team worked to include covariate data such as COVID-19 cases, mobility data, government actions such as shelter-in-place orders, and online traffic. Once we became familiar with the  client’s product range, it became clear we  needed to split products into 2 buckets: those with short lead times and those with longer lead times, as different forecasting approaches and datastreams were relevant for each bucket.


Forecasting is a particularly powerful tool for ecommerce businesses because the space allows for  access to a plethora of data and data streams - fantastic for some of the data hungry algorithms in the Remi AI Forecasting Engine. In the case of this client, a combination of quality data, powerful technology, and a strong project team produced the accuracy results that the client needed to succeed.

The Solution

A North American client  that was already experiencing rapid growth found it accentuated by the COVID-19 pandemic of 2020. They believed  advanced demand forecasting would be of great benefit to their operations and solve two core problems: Lengthy lead times and  volatile demand. This combination had proved to be a challenge for traditional forecasting methods, leading to a large number of products that were overstocked with more than 26 weeks of supply.



Long lead times coupled with a  rapid shift to online during COVID-19

Automated Machine Learning Demand Forecasting utilising many covariate data streams

Demand forecasting Ecommerce.png
bottom of page