Making Predictions on Warranty Claims

[McKofi]

Please help.

How do I analyze warranty data using mileage traveled to make reliability predictions on future claims?

I have no information how long this product was in used. Specifically, all I know is the miles traveled before warranty claim was made received.

Thanks.

[adam]

You can do the analysis based on miles. Instead of having time-to-failure, your random variable is miles-to-failure. Your results will be the probability of failure vs mileage (or percent failures at different mileage intervals). The only concern is the suspensions. You will also need to include the suspension in terms of mileage. There are different ways of estimating the mileage on the suspensions. The easiest will be to use an average mileage per unit time (e.g., per month) and based on the sales date of the item estimate its mileage.

There are some more involved methods such as estimating the mileage (usage) distribution and applying this distribution to your suspended population. The difficulty in this method is obtaining this usage information. One approach is to use the failures. By taking the mileage of each failure since its sales date you can figure-out the mileage per unit time. You can then fit a distribution to these miles per unit time and therefore obtain your usage distribution. The danger here is that you may get a skewed distribution because it may be that all the higher usage customers are the ones that experience most of the failures.