It looks like you want to know how well your PROCESS is working, which means
you want to know how many errors are you making versus how many
opportunities for errors exist. For this the DPMO calculation is the best
indicator, but it means you have to calculate the opportunities for errors
for each and all boards. With some work that become easy enough, because
somehow that information is "hidden" in the data files for the boards you
are building. And if your process is really good you are making fewer than
50 errors per million opportunities.

On the other hand, if you are in Marketing, you don't care about the
problems the guys on the production line are experiencing, you want to know
what your customer is going to see. If you as a customer just bought the
fanciest pick-and-place machine, with all the possible bells and whistles,
and it does not work, all you are going to think is: 100% failure and you
are going to let the manufacturer know!! Even if it has millions
opportunities for failures built in!!

In between you have First Pass Yield, which to most people means how many
boards that come of the line work versus how many boards did you make.
Really, an indicator good for nothing.

And remember, statistics don't mean much if you use hardly any data points.
(If all you do this week is solder one resistor in place and one of the two
joints fails you have a DPMO of 500,000, maybe 330,000 if you count the part
itself too.)

When it comes to corrective action, look at the classes (choose 7 to 10
classes) of errors you are making and analyze the most prevalent class,  or
the class that costs you most in time or parts, first. (Pareto charts.) Also
remember, if you take too much time collecting, sorting, analyzing, etc the
data, the problem will by then be so large that .............(fill in any
very bad consequence.) Therefore you need a system that is easy to work with
and that allows you to get to an answer quickly.

Have fun,
Ahne.




-----Original Message-----
From:   TechNet [mailto:[log in to unmask]] On Behalf Of Paul Signorelli
Sent:   Friday, December 27, 2002 10:09
To:     [log in to unmask]
Subject:        [TN] Normalizing Yield Percentages ?

Does anyone have a standard way to normalize yield percentage charts for
wildly varying production volumes ?
My yield percentage charts show 50% yield when there is 1 out of 2
failures or 50 out of 100 failures.
The 50% Yield with 1 out of 2 usually happens during a Christmas week
when there is no production, only rework.
I have tried to normalize the charts by using a constant denominator
based on the nominal volume.
I use 13 week rolling averages but they are messed up by the low volume
yield results.
This low volume percentage reporting can cause incorrect corrective
actions.
Does anyone know of a standard way to normalize the data ?
Cronbach's Alpha,
Split-Half Reliability (Spearman-Brown Coefficient),
Attenuation correction,
Pair Wise Deletion of Missing Data,
Partial correlation,
Parzen Window,
Pearson Curves,
Johnson Curves ?

Paul Signorelli

---------------------------------------------------
Technet Mail List provided as a free service by IPC using LISTSERV 1.8e
To unsubscribe, send a message to [log in to unmask] with following text in
the BODY (NOT the subject field): SIGNOFF Technet
To temporarily halt or (re-start) delivery of Technet send e-mail to
[log in to unmask]: SET Technet NOMAIL or (MAIL)
To receive ONE mailing per day of all the posts: send e-mail to
[log in to unmask]: SET Technet Digest
Search the archives of previous posts at: http://listserv.ipc.org/archives
Please visit IPC web site http://www.ipc.org/html/forum.htm for additional
information, or contact Keach Sasamori at [log in to unmask] or 847-509-9700
ext.5315
-----------------------------------------------------

---------------------------------------------------
Technet Mail List provided as a free service by IPC using LISTSERV 1.8e
To unsubscribe, send a message to [log in to unmask] with following text in
the BODY (NOT the subject field): SIGNOFF Technet
To temporarily halt or (re-start) delivery of Technet send e-mail to [log in to unmask]: SET Technet NOMAIL or (MAIL)
To receive ONE mailing per day of all the posts: send e-mail to [log in to unmask]: SET Technet Digest
Search the archives of previous posts at: http://listserv.ipc.org/archives
Please visit IPC web site http://www.ipc.org/html/forum.htm for additional
information, or contact Keach Sasamori at [log in to unmask] or 847-509-9700 ext.5315
-----------------------------------------------------