Squeeze in one more - Rate of failure

Candi1024

Quite Involved in Discussions
I am attempting to determine a historical rate of failure, or some kind of metric that I can use to measure a certain failure rate.

We have a database of returns to customer service and reasons. We have production database for any given time.

The arguement is that the returns last year does not directly relate to the production last year, because it would have been items that were produced years before that were returned.

Some mentioned the amount of failures over 1 year by how many are in the field. I don't like the way that takes only a part of failure over the total produced.

Some mentioned the number of returns that had this particular failure. I don't like that number because it doesn't take into account how many produced.

Being that we produce approx 5k a year, and the returns for the particular failure were approx 150 last year, even considering advanced age and customer damage, I can show some reliability. Do you agree with this, and how can I convice others?
 
Last edited:

Mike S.

Happy to be Alive
Trusted Information Resource
Any serialization or way to determine how long a unit was in the field before it failed or was returned?
 

Candi1024

Quite Involved in Discussions
However as far as I know, there is no metric or guideline for mean time of failure.

I suppose I can go back 5 years and see how many produced were returned.
That's a lot of research! lol
 

Bev D

Heretical Statistician
Leader
Super Moderator
The standard way to trend this would be either MTBF or cumulative failure rate grouped by serial number 'batches' or production month.


I have attached an article explaining the approach...
hope it helps.
 

Attachments

  • Warranty Cumulative Frequency Trending.pdf
    67.5 KB · Views: 176

Candi1024

Quite Involved in Discussions
Nice white paper. I'm saving that ;)

I found a way to search our returns by year produced, no matter when the return occured. I am then planning to compare that number to the number produced in the same year.
 

Miner

Forum Moderator
Leader
Admin
I haven't had an opportunity to read the white paper, so this may have been addressed. If your product sits in inventory for a long time before use, or if it runs a relatively long time before failure, you will see a downward trend in the more recent time periods. You should collect your failure rates from the stable time periods prior to this downward trend.
 
Top Bottom