Measuring supplier quality with small quantities

Tagin

Trusted Information Resource
Any recommendations for monitoring supplier quality when the monthly quantities can sometimes be small (e.g., qty 10-20)?
  • If we have 1 failure our of 100, ok, that's 1%.
  • But then if we have a month with 1 failure out of 10, it's 10%; but that 10% could be misleading since qty 10 is a very small "sample size".
It makes it difficult to get any sense of quality trends.

The supplied components are computer parts. We are ISO 9001, not aerospace or automotive.

Any recommendations?
 

jam325

Starting to get Involved
You could use Acceptable Quality Level sampling. It gives you a number of units to sample based on the size of a lot received. Then based on the number of pass and fail units you decide to accept or reject the lot. Then you could monitor the supplier based on the number of accepted and rejected lots.

Here is a thread that talks about it on Elsmar but it won't let me post it.

These sampling plans tend to be used for incoming inspection and can tend toward rejecting material when you find a single bad unit in a small lot. So you have to be careful not to start rejecting every single lot that you receive.

Hope that helps!
 

Marc

Fully vaccinated are you?
Leader
Here is a thread that talks about it on Elsmar but it won't let me post it.
It takes 10 posts before a person can post an active link, but technically you can remove the https:// elsmar.com part and post the rest. If you do I'll make the link active. I do watch for situations like this.
 

Tagin

Trusted Information Resource
The failures we detect are in Production, not receiving. Due to the complexity of the components, it is impractical to test them prior to Production.

Also, rejecting a lot is not feasible due to the high demand for the product, and short supply of the components.

I suppose maybe I could do some kind of AQL-like analysis ex post facto, but it still suffers from the issue of small and variable quantities.
 

Ninja

Looking for Reality
Trusted Information Resource
In your situation, it seems more effective to arrange a return/replace relationship with the vendor rather than to apply high volume metrics to a low volume process...

Defects as % may not be very useful (as you describe).
What about defects/month, per year, per last rolling hundred...etc?
 

Tagin

Trusted Information Resource
In your situation, it seems more effective to arrange a return/replace relationship with the vendor rather than to apply high volume metrics to a low volume process...

Defects as % may not be very useful (as you describe).
What about defects/month, per year, per last rolling hundred...etc?

"Last rolling 100"...that's an interesting idea. Hmmm, it could mean we have 4-6 month delays in initial data (e.g., if we consume ~20/mo), but then as long as the ongoing runrate has some consistency this might work. Of course, it also means the "responsiveness" of the numbers to a sudden change in quality is reduced, but that is part of the intent. And after all, just having this data doesn't mean we will blithely ignore what happens on the floor daily/weekly.

I'm going to see if I can do some test runs with this kind of approach - thanks!
 

Golfman25

Trusted Information Resource
I wouldn't bother. You need a minimum sample size to be statistically relevant. You could just track "incidents" to keep them on the radar.
 

Bev D

Heretical Statistician
Leader
Super Moderator
:unsure: a simple categorical data control chart will do the trick. you could try a p chart that will give you variable limits to adjust for 'sample size'. or you might be better off with an I, MR chart. we'd have to see the data to know.

AQL and other 'statistical' approaches are not appropriate.
 

Ninja

Looking for Reality
Trusted Information Resource
FWIW...many customers give a "supplier scorecard" or similar once per year to select vendors.
I do not suggest that you do this for your small volume things...but it gives some guidance for a "per year" approach.

I understand that you're trying to rate a supplier in a general 'across the board' way...but I would simply look at their defects per year and how much the return/replace process costed for that year...and measure them by dollars cost due to defects...then use that to negotiate the next year's prices...

Small volume: target 'effective' instead of 'standardized'...my two cents...

And after all, just having this data doesn't mean we will blithely ignore what happens on the floor daily/weekly.

That's what's really important...out of curiosity...why are you trying to do more (beyond tracking cause of the floor disruptions).
 

Mike S.

Happy to be Alive
Trusted Information Resource
Another option..... unless this supplier supplies a critical, strategic part and the effect of even 1 bad part from them is significant (doesn't sound that way) maybe you don't formally measure small volume suppliers like this at all. Focus first on the 20 % of suppliers who probably supply 80% of the product (and maybe 80% of the problems) and only after you do all you can on these big boys do you put much effort into the next teir down.

You can't give every problem equal attention.
 
Top Bottom