Forecasting and collaborative forecasting
You have seen that ERP systems can support many activities in the supply chain. Also, for key activities collaboration between actors in the supply chain is needed that is not always fully supported by the ERP systems. In these cases special purpose systems are acquired or developed and human expertise is needed.
In this article we look at the forecasting process as an example. We use the overview of Pennings and van Dalen on collaborative forecasting and the work of the the GS1 Collaborative Planning, Forecasting & Replenishment Workgroup (CPFR®) Workgroup (both references are stated below).
Retailers use a forecast as input for sales, inventory and order decisions, suppliers for production and procurement decisions, and distributors for capacity allocation decisions. The forecast horizons and units are different for these forecasts: retailers have to forecast consumer demand for short-term stocking decisions, whereas manufacturers have to forecast the actual orders placed by the retailer for long-term production decisions. Errors in a forecast can propagate upstream, distorting the basis on which decisions are made. Even if no errors are made, demand variability increases upwards in the chain – the bullwhip effect.
If all parties in the supply chain develop their forecasts in isolation, they do not make use of the knowledge their supply chain partners have. Sharing information such as sales information, market conditions, stock levels and lead times can make plans more accurate. In addition to sharing information, supply chain parties can also share knowledge and capabilities in working together. However trust is needed between supply chain partners which is not always the case in highly competitive markets.
Sales promotions may have a huge impact on demand. Many other factors may play a role such as product reviews, consumer opinion, regulation, product launches or price setting by the competition etc. As many of these factors can be tacit it is often not feasible to build fully automated forecasting models. Human expert judgement plays a role in such cases.
A typical ‘modern’ forecast process uses Industry information, Historical Information and Sales information. It combines these data sources with information from Sales Directors, product planners and operation managers. The information is used to create various kinds of forecasts using judgmental and statistical techniques also taking promotion plans into account. Based on consensus an ultimate forecast is developed to use as input for planning. Popular techniques to calculate forecasts include regression analysis and simple exponential smoothing (SES). Forecast errors can be intentional or unintentional. Intentional errors can be a result of company politics or competitive forces.
CPFR was setup as a reference model to combine the intelligence of multiple trading partners in the planning and fulfilment of customer demand. CPFR links sales and marketing best practices, such as category management, to supply chain planning and execution processes to increase availability while reducing inventory, transportation and logistics costs. It provides a framework for supply chain partners to collaborate. Key area’s to share information and knowledge and work together are shown in the figure above: Joint planning includes the exchange of strategic goals and plans. In the forecasting process sales data and market analysis are shared. Production and distribution are jointly setup in the execution phase. Finally, exceptions are jointly managed and performance is evaluated together.
Collaborative Forecasting by C.L.P. Pennings and J. van Dalen In: Cross-chain collaboration in the fast moving consumer goods supply chain: Kok, Ton de and Dalen, Jan van and Hillegersberg, Jos van (2015) Cross-chain collaboration in the fast moving consumer goods supply chain. University of Eindhoven. ISBN 9789038638140 (Open Access)
GS1 - Collaborative Planning, Forecasting & Replenishment Workgroup - CPFR Guidelines and Resources
© University of Twente