Issue 49, March 2005

Using Warranty Return Data to Compare Different Product Designs

Warranty data analysis is crucial for predicting the quantity of products that will be returned under warranty. Warranty data analysis conducted using Weibull++ can allow an organization to make the most efficient allocation of resources to warranty services provision. This allows you to anticipate customer support needs and take the necessary steps to insure customer satisfaction with the warranty process. Warranty data analysis helps in quantifying the number of spare parts or replacement units to stock or supply to distributors. It also helps in scheduling and planning for service, shipping and distribution personnel and in budgeting for all the warranty service needs.

Warranty data analysis can also be performed to compare different designs of a product. By comparing return data of different versions of a product (different designs or designs containing components from different suppliers), you can get a clear picture of the number of returns to expect depending on the version (design) of the product and can make a more informed decision about which design is more reliable.

In this article, we study a situation in which a manufacturer had been producing a product and then incorporated a redesign in the product in an attempt to reduce failures in the field. By analyzing the warranty returns of each design, the manufacturer can confirm whether the new design is, in fact, a more reliable design, and assess the cost savings that resulted from switching to the new design.

Weibull++ Warranty Analysis

The manufacturer collected 6 months of returns data for the product under the old design (Design A):

Figure 1: Warranty returns data for Design A (old design)

The previous table is usually referred to as a "Nevada warranty chart." This is the failure data entry format most commonly used in industry and is the format provided in the Warranty Analysis tool of Weibull++. This tool does not introduce any new analysis methodologies, but rather automates a process which would normally be quite time-consuming and tedious. The table can be interpreted as follows:

The number of units sold/shipped at the beginning of Jan-04 is 400, the number of units sold/shipped at the beginning of Feb-04 is 420, and so on. The number of failures obtained from the Jan-04 batch is 2 for Jan-04 (i.e. at the end of Jan-04, or 1 month after being put in service), 3 for Feb-04 (i.e. 2 months after being put in service), etc.

Using this data, a reliability distribution model can be obtained. The model is as follows:

Figure 2: Model parameters for Design A (old design)

At the beginning of July 2004, the manufacturer switched to a new design (Design B). The following table is the Nevada chart for Design B's warranty returns.

Figure 3: Warranty returns data for Design B (new design)

The obtained model is as follows:

Figure 4: Model parameters for Design B (new design)

Expected Warranty Returns Comparison Analysis
To further assess the improvements that have been incorporated in the new design, the manufacturer can compare the expected warranty returns for Design A and Design B.

For the following 6 months, the manufacturer expects to ship/sell the following number of units:
 

Figure 5: Sales forecasts for the next 6 months

Based on these sales forecasts, the manufacturer can estimate the number of returns for Design B, and for Design A if it had not been replaced by the new design (i.e. for Design A, the above sales figures would be for July 04 to December 04).

The expected failure returns can be estimated, using the reliability distributions and the sales units, by clicking the Generate Forecast button. The results are as follows:
 

Figure 6: Design A's failure forecasts for next 6 months

Figure 7: Design B's failure forecasts for next 6 months

The next figure shows a comparison of the different designs failure returns. In order to avoid including failures that would occur in the units already in the market (because the number of each product version sold each month was not the same for the two versions), only the returns predicted to result from future sales are analyzed.

Figure 8: Future failure returns for the two product designs

Figure 9: Comparison of the future failure returns for the two product designs

The manufacturer estimates that each warranty failure return costs the company $120. The following plot shows a monetary comparison of the two designs.

Figure 10: Comparison of the future failure returns costs for the two product designs

The total number of returns under warranty for Design A over the next 6 months periods is estimated to be 230 units, resulting in $27,600 of warranty replacement costs. Design Bs estimated total number of returns under warranty is 154 units, resulting in $18,480 of warranty replacement costs.
 

Copyright 2005 ReliaSoft Corporation, ALL RIGHTS RESERVED