Medion AG, a manufacturer of consumer electronics and information technology products, has established its reputation through excellent after-sales service. In the first part of this series, we looked at Medion's challenges in forecasting demand for replacement parts. (See: Creating an Optimal Parts Management System, Part 1.) In this blog, we look at how it developed an innovative system to manage replacement part needs.
With the help of the consulting firm Abels & Kemmner, Medion evaluated consumption values for replacement parts. Because of their short consumption periods (less than six months to a year) and unsteady demand, classic methods such as the mean average, exponential smoothing, and median were not suitable for forecasting and had no advantages over the methods Medion had been using. Therefore, past projects were looked at more closely, and consumption values were analyzed with regard to failure behavior.
Patterns emerged in the process. Five different types of failure patterns were identified and assigned what are known as standard curves. The components were assigned to the curves depending on their failure rates in the initial use period, in the middle, and at the end of the warranty period. It came as no surprise that the patterns were not necessarily assembly-specific. Electronics components fell into all of them, and so did mechanical components, so the standard curves could not simply be assigned by product group. Particular rules and indicators had to be used instead.
Methods were then developed to use these standard curves to forecast demand for replacement parts. A prototype of an analytical software tool was specially developed for this very complex task involving lots of calculations. The failure data was processed, and forecasts were calculated.
By assessing the first phase of the consumption period, users of the tool can assign the replacement part to the most suitable failure pattern and calculate demand expected during the warranty period. Options were also built in allowing users to adapt forecasts according to experience-based knowledge. The forecast values can be stretched or compressed by simple clicks, and consumption periods can be reweighted. The tool is very chart-oriented and provides information very quickly on consumption and demand.
The final step saw the methodology being validated. Forecasts based on information known at that time were calculated for certain times in the past and compared with forecasts generated using previous methods. The result was clear — the new methodology makes it possible to make better forecasts for replacement part demand. In particular, the cover for residual demand was improved by a double-digit percentage, enhancing service while reducing stocks. This project was crowned as a double success.
The new methodology has been implemented in the standard DISKOVER SCO software to optimize forecasting and scheduling, so the relevant functions can be used for replacement part management.