Analysis
Message posted by Matt (via 198.26.130.38) on November 1, 2001 at 9:28 AM (ET)

I am in the military and work at a repair facility for avionics. I have been tasked with determining what our "failure rate" is for the command. The "failure rate" being how often a particular type of avionics gear is returned because it doesn't work after we have repaired it. We work on over 700 different pieces of gear. Can I get an accurate "failure rate" by dividing the the number of returned pieces by the number of non returned pieces? Or do I need to break it down by each of the different types of gear. There are some pieces of gear that we may only repair 6 a year, there are others that we may repair 6 a day. What determines the threshold for even looking at some data?
Thanks in advance,
Matt

READERS RESPOND:
(In chronological order. Most recent at the bottom.)

Re: Analysis
Message posted by Tomi (via 154.32.142.50) on November 4, 2001 at 3:46 AM (ET)

However, a more complicated failure rate (taking into account the type of gear) might be better for your situation. The use of a weighted mean might help to distinguish between trivial returns and more significant ones.

I guess the main thing is that this failure rate should be applied consistently to compare failure rates over the years. However it mustn't be too complicated.

Do you have an inventory system at the moment? I would suggest you build a database of repaired items and returned items and build in an automatic calculation. This would be a serious project but probably worth it in the long run. You could build in the capability to calculate an overall failure rate and individual failure rates (allowing you to focus on areas that need improving).

If you want further help with this, please email me at tomi@hautlieu.sch.je