RE: Utilizing Process Behavior Chart to Forecast Weekly Shipments

[Rip Stauffer <rstauffer@bluefirepartners.com>]

Dan,

Your approach seems reasonable. I'm not sure of what algorithm your new IT
Manager is using, but you should get him to explain it. There are some good
methods for decomposing data with strong seasonal components, but in the
absence of any special cause signals in your control charts, time-series
trends are probably not significant enough to be real; that is to say, there
is probably no cause associated. 
Do these "trends" actually show up in some predictable periodicity? For
example, is every July better or worse than every June? If so, there may be
some reason to suspect some seasonality. Your description of the process
behavior charts suggests otherwise, however.

You might want to examine Don Wheeler's new book, "Making Sense of Data." It
is an expansion of the Wheeler and Poling "Building Continual Improvement."
It has some pretty good stuff on seasonality and trended control limits, and
may be an excellent gift for your IT manager, if he/she is inclined to read
and learn from new material. The book is available from SPC Press, Inc.,
ISBN 0-945320-61-2. 

A deeper question I have, though, is whether you are measuring the right
thing. You might want to construct some charts dealing with your
forecasting; from what you've said, that might be the more interesting
measurement. Some upstream measures would also be valuable--if you can
identify them--in predicting what the output will need to be.

I'd love to hear from Steve Prevette on this one...he has done some very
interesting things with trending. Also, Davis Balestracci has done a number
of things, as well, but I don't know whether he has published a lot of them.


Best regards to all,

Rip

Rip Stauffer, Senior Consultant
BlueFire Partners
1300 Fifth St. Towers, 150 So. Fifth St.
Minneapolis, MN 55402
612-344-1027
mailto:rstauffer@bluefirepartners.com
http://www.bluefirepartners.com/