Data With Seasonality

[Paul J]

The number of product returns I receive is strongly seasonal. Perhaps my product is used more in the summer and/or my customers are more likely to return the product for service in the summer. Does a time-based life analysis (e.g. using the Nevada Analysis feature in Weibull ++) make sense?

[Arai.M]

Paul,

Unless you have a clear understanding of usage and can apply a specific usage profile to your failure data and suspensions, I can't think of many alternatives. Using a Nevada Chart will assume a homogeneous usage. This might not be too bad of an approach if the granularity needed of the results is not down to the month; for example if the expected returns for a year are good enough.

An analysis in terms of usage might be a bit more accurate. One example is using the Usage-Based Warranty Analysis http://www.weibull.com/hotwire/issue73/hottopics73.htm (http://www.weibull.com/hotwire/issue73/hottopics73.htm). You have to be careful on how the suspensions are treated here (the usage is assumed to follow a distribution but not a pattern as you might see with seasonality). It can however give you an idea of how to analyze data if usage can be defined.

Another way of looking at this is with stress profiles within the context of accelerated life testing. An article that might address this can be found at http://www.reliabilitynews.com/v8i2/warranty.htm (http://www.reliabilitynews.com/v8i2/warranty.htm).

Best regards,
Arai