Problem
A manufacturing company produces a large number of articles of different types. This includes articles with regular and seasonal as well as articles with intermittent demand. New articles as well as top-sellers, both gradually increasing/decreasing and stable products are part of the portfolio. How many of which article group will be sold in the next few weeks?
Solution
An automated forecasting system is developed which provides forecasts
on article group level for the next couple of
weeks and thus satisfies the requirements of different types of planning objects.
Methods from statistics as well as machine learning and pattern recognition methods are used to meet the challenges of regular as well as intermittent demand patterns.
Existing open orders are taken into account in the forecast models.
A classification mechanism determines which article groups can be predicted well by artificial intelligence, and for which planning objects additional validation by experts is recommended.
An analysis on different hierarchical levels determines whether the article groups are preferably forecasted at the current level or instead by top-down or bottom-up mechanisms.
The data is taken from the company’s ERP system and final results are fed back into it.
The company uses its own reporting tools for
displaying and visualizing the results.
A dashboard shows intermediate results for planners and provides expert information
for data scientists.
Benefits
- Point out the quality of forecasts of different article groups to production planners; suggest which of them still require manual planning input
- Reduce storage and depreciation costs
- Ensure delivery capability
- Save time for manual planning
Download
You can download a detailed use case on the topic “Supply Chain” in German:
Download Supply Chain case study (German)
You can find out more about safety stocks and the optimization of warehousing in this blog post (in German):
Optimize safety stocks
