Analysis of Data to Predict Warranty Cost for Various Regions

Nowadays, automobile manufacturers are assembling vehicles at one location or multiple locations and selling them in the global markets. The Original Equipment Manufacturers (OEM) are asking the suppliers to share the warranty return cost for their returned components. This recent phenomenon is putting the supplier in precarious position, financially, by sharing the warranty cost when supplier margins are very thin. It is creating a challenge to suppliers how to predict the warranty cost based on components sold and forecasted to sell in coming months and years when vehicles are selling world wide carrying different warranty terms. Typically, luxury vehicles sold in North America carries warranty period of 4 years or 48,000 miles whichever comes first. In Asia, Australia, Europe and Africa warranty period is 2 years irrespective of mileage on the vehicle. Vehicle sale volumes are going to change based on currency fluctuations and demand for the particular product. This poses a challenge to the supplier's in predicting the warranty costs for various regions. This study shows that, how to take the complex failure information for different regions and break it down to main categories. The vehicle sale information has been divided into North America and rest of the World (ROW). In the same way, failure data related to particular system warranty separated by regions. ROW market failure data has been analyzed by using the Reliasoft Weibull ++ software with forecasted quantities using time as univariate. Two variables were identified to calculate the warranty forecast for the North American market. One is time and another one is mileage on the vehicle. From our experience, the failure rate of a sub system is more dependent on usage rather than time. The data clearly shows that, customers are putting 40 miles per day on average and 80% of vehicles are reaching the warranty usage limit before the time limit. Based on this data, and failure rate is dependent on usage rather than time. It is easy to do calculations by employing the univariate plan rather than complex analysis by taking time and usage parameters. Warranty costs are calculated based on Maximum likelihood estimation. This approach is relatively easy to follow for small to medium suppliers in the automotive industry.