Recording Real Time to Failure

If the mean time to failure is an important statistic, why only measure it during development and testing? Building a small, anonymous crash reporting tool into your application can provide you with a wealth of information and can help you to improve your overall quality. Of course it would be wise to make sure the users are aware of this tool, wiser to give them the ability to enable and disable it as they see fit, and wisest to leave it disabled by default.

The crucial piece of information here is the time to failure of the application. For this it might not be a bad idea to capture the time at which the application was started and the time at which the application failed. From those you can calculate how long the software ran but at the same time you have a little more detail at your disposal. It would be interesting to see a system in which crashes always occurred at the same time of day.

It would also be very useful to capture information about the hardware and operating systems being used to run the application to perform an analysis of the failure rate across different platforms. It’s always possible that there is a fault which is only triggered on a certain hardware configuration, and this would enable us to pinpoint these types of issues.

Information reported by this tool could be kept in a database and used for even more analysis over time. For example, calculating the mean time to failure grouped by month will give an idea of whether the stability of the product increases or decreases as we work to repair problems and improve quality.