Demand management & statistical forecast

Industry
€5 bln semiconductors manufacturer with >30k employees, a large product portfolio, with manufacturing, R&D and Sales operations across the globe.

Challenge
Forecasting is an essential process in the semiconductor industry as overall lead times are significantly longer than the required lead times from customers. Having an accurate forecast that is based on an efficient process and provides opportunities to quickly evaluate various scenarios is the challenge for this company. EyeOn supported to make significant progress in all of these areas.

Project
In order to discover what improvements are required a quantitative and qualitative forecast assessment was conducted via interviews and various interactive workshops. Already a history of organizational changes and statistical forecast adjustments were made in the past. Topics that have been scrutinized are a.o. how to measure forecast accuracy. As forecast accuracy is the base to measure improvements it is an essential part to have the right set-up chosen. Important is to decide on the right data (e.g. billings, bookings, shipments), the level of detail to measure (SKU, type, family etc.), horizon and time lag.

As statistical forecast can provide a lot of efficiency and accuracy gains it has to be investigated for what areas this applies. A very important aspect of applying statistical forecast I to understand when it adds value when it does not. In this case a quantitative evaluation was done to measure in what areas the manually created forecast performed better. This understanding led to the creation of a table that made clear when to apply statistics and when to do a manual forecast and by who. Important aspects that were taken into account were demand variability, sales value and supply risk based on the number of customers related to the sales. An important aspect was to deviate from the starting point of the ABC-XYZ analysis and to add a third dimension being the supply risk as expressed by the number of customers in three categories.

In order to use the new set-up the complete forecast process required a redesign. Important aspects were central vs decentral roles and responsibilities, process efficiency measurements via time spent, setting up PDCA cycles, adjustments in the technical statistical forecast approach. The whole change required serious change management approach which is always part of the EyeOn approach.  

Results
From 800+ to < 60 forecasters whilst improving forecast accuracy by 10%. Implemented a three dimensional forecast framework to make clear when to apply statistics. Redesigned and re-implemented the forecast process.