Data considerations for Warranty Prediction

The data should indicate the number of items shipped and returned in each period
The data should track the number of systems shipped and the number of systems returned from the shipment in subsequent time periods. For example, an automobile company tracks the total number of cars shipped each year, and the number of cars returned each year from each annual shipment.
Your data should be in arbitrarily-censored format
Arbitrarily-censored data is arranged in intervals of time (or other usage unit) and indicates the number of items that fail within each interval. Therefore, your data should be arranged in columns that contain the start time and end time of each interval. For exact failure data, the start time and end time of the interval are the same (the exact failure time.) For right-censored data, the end time of the interval is a missing value (*).
Warranty prediction does not use left-censored data. If you enter the start time of the interval as a missing value (*), you must also enter the end time interval as a missing value.
If your data are in triangular matrix form, which is commonly used to record raw data for warranty claims, you must first reformat your data into arbitrarily-censored time-to-failure data before you perform this analysis. For more information, go to Pre-Process Warranty Data
The distribution that you select must fit your data adequately
If the selected distribution does not fit your data well, the estimates of the expected number of failures will not be accurate. To determine which parametric distribution best fits your data, use Distribution ID Plot (Arbitrary Censoring).
By using this site you agree to the use of cookies for analytics and personalized content.  Read our policy