ToolsGroup Embeds Machine Learning Technology in Demand Forecasting

ToolsGroup, a provider of demand analytics and supply chain planning (SCP) software, claims to be the first vendor in its class to embed machine learning into a commercially available demand forecasting product. The new technology has been proven to significantly improve forecast accuracy in major global customer installations. ToolsGroup's embedded machine learning is available as part of the SO99+ version 7.3.1 demand forecasting solution (“SO99+” stands for “service optimizer over 99%”).
Forecast accuracy is an important driver of business outcomes, yet most demand forecasting systems today often produce disappointing results and significant forecast errors. The traditional standard models found in these systems cannot easily identify trends in the data. Their inability to model the underlying causes of demand variability can also lead to poor trade promotion execution and failed new product launches. Last but not least, they can be manually intensive, resulting in poor planner productivity.
More than three years ago ToolsGroup saw that being able to take advantage of more data (downstream data, market data, web data, social media, etc.) in the forecasting process could drive improved forecasting, inventories, and supply chain performance. It was clear that conventional forecasting and modeling techniques could not scale to fully take advantage of this data. Therefore, the vendor developed and started deploying a machine learning engine (MLE) as an add-on to its standard product, and has since deployed about 10 implementations.
ToolsGroup has harnessed the power of machine learning to more accurately model demand in difficult forecasting scenarios such as trade promotions, new product introductions (NPIs), extreme seasonality, and product cannibalization. Machine learning takes demand forecasting to new heights because older technologies could not solve the difficult problem of measuring the impact of external stimuli on baseline demand. ToolsGroup’s demand modeling creates a reliable baseline, then uses machine learning to adjust the baseline by identifying the effect of stimuli and demand indicators at a detailed channel level. It analyzes all the relevant variables and the complex interactions among them in an automated fashion.
As ToolsGroup gained experience with the technology, it was able to decrease the footprint and increase the processing speed, reliability, and product standardization. It can now embed the solution in its standard product, with the implications and benefits described in the aforementioned release. The vendor’s renowned consumer goods clients cite results such as notable reductions in forecast errors and in lost sales, improved inventory mix, and payback improvements in promotions. Other ToolsGroup customers are using machine learning for NPIs to identify which new products will have significantly above average sales. Identifying those successful new products early on allows companies to determine the most advantageous way to allocate additional marketing, purchasing and replenishment resources to up and coming products.
For more background see TEC’s previous blog post on ToolsGroup.
comments powered by Disqus