Collaborative Forecasting: Goodyear Tire & Rubber Company's Journey

Describes in detail the problems in the legacy system at Goodyear and the changes the company made over time to improve it ... the new forecasting system emphasizes customer specific forecasting ... for new product forecasting, Goodyear uses an analog model. The Goodyear Tire & Rubber Company (Goodyear) is a worldwide leader in the production of tires, engineered products and other goods, with over one hundred years of history. It has long recognized the importance of forecasting customer demand as a part of its production and distribution planning, and the need for meeting the customer's requirements with the right product in the right place at the right time, every time. Statistical forecasting techniques have long been used to begin the process of negotiation between production planning and marketing organizations. This vital process starts with the need for forecasting the short term unconstrained demand for developing production and distribution plans, which is the core of our demand planning business model. Goodyear has made large improvements in the forecasting process, systems and accuracy, despite the various business complexities it experienced in the last decade or so, which are: 1. Inclusion of our Kelly-Springfield subsidiary into the Goodyear forecasting system 2. Acquisition of Dunlop Tire 3. Need for specific customer forecasts 4. Continuous introduction of new and innovative products, which expanded our product lines In spite of these dramatic changes, the Goodyear branded products sold to our dealers and company owned outlets have remained a critical portion of our business. Considering this segment of business as a control group, we made a tremendous progress in our forecasting efforts, reducing forecasting error at a SKU level from 65% in 1993 to 31% in 2003. Forecasts were prepared 60 days ahead, and were weighted by volume for computing average percent error. FORECASTING SYSTEM IN HISTORICAL PERSPECTIVE The forecasting system at Goodyear was started in early 1970's with the installation of IBM's IMPACT mainframe software at our North American operations, which used time series models. Shortly, thereafter, our Kelly-Springfield subsidiary developed its own forecasting software solution. In the late 1980's, the North American Tire Division partnered with outside consultants to develop a new forecasting system that used sophisticated models, such as Holt triple exponential smoothing model. The system collected order information from our ERP system and then used the SAS program to develop forecasts. By early 1990's, Goodyear had three distinct forecasting systems with independent consensus processes, different metrics, and no electronic communication amongst them. Figure 1 describes the system's landscape, which existed at the time. IMPROVEMENTS IN OUR LEGACY SYSTEM Within the Goodyear US system, we were experiencing extremely large errors at a SKU level (both at item and location), when forecasts were prepared 60 days ahead. This clashed with our longstanding policy of providing our customers the highest level of customer service and fulfilling orders when requested. So, we embarked on a project to increase the forecast accuracy. Since we found that our system was doing a good job with the Holt's triple exponential smoothing technique at an aggregate level, we replaced the linear regression models used for SKU level forecasts with single exponential smoothing model. With that, we immediately experienced 20% improvement in our forecast accuracy, weighted by volume. In 1993, the North American Tire division (which included US and Canadian operations) made a decision to implement SAP R/2 as our ERP solution. With the result, the forecasting systems in US and Canada were revamped. We used this opportunity to combine the US and Canadian forecasting systems into one integrated application for Goodyear. …

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.