Companies in the manufacturing and commercial industries are particularly reliant on the timely availability of goods on the market. It is therefore important to be able to forecast the future demand from customers as closely as possible.
The forecasting models implemented in the ERP systems of well-known software manufacturers rarely produce really good forecasts. Furthermore, to ensure that forecasts are of a usable quality, too many parameters have to be monitored for each and every item. The result is that many companies do not use automatic demand forecasting and assign planning tasks to their specialist staff.
Avantgarde has developed an adaptive forecasting method as part of a Master dissertation in cooperation with Hochschule Niederrhein University of Applied Sciences. The idea behind this project was to minimise the workload incurred in manually updating the forecasting parameters by implementing a "self-learning" algorithm.
The algorithm allows data to be rendered anonymous to a high degree and has yielded good forecast results. Users also have the possibility to include additional parameters in the forecast, in addition to the usual data templates (trend, seasonal, etc.). Companies are now able to check product demand automatically for dependences on other statistical time series. Initial tests are done anew for each period.
"Our first tests on reference projects have showed highly promising results," said Thomas Müller, General Manager of Avantgarde.
Prof. Stegemerten was also very pleased with the quality and the results of the project: "Innovative solutions come from innovative approaches. Our cooperation with Avantgarde has yielded excellent results once again. Combining existing mathematical methods with an innovative concept can lay the foundations for new products. This is precisely the kind of thing we hope to achieve in such cooperative projects."