M
I am an intern and have been tasked with developing statistical process control charts for a number of processes at my place of employ. Unfortunately, I will not have taken a class on statistical process control until this fall; furthermore none of the industrial engineers here know how to make a control chart. Therefore I have turned to searching the internet, I've found answers to the majority of my problems, but I'm having issues establishing useful upper and lower control limits, as the data is wildly inconsistent.
I have been instructed to use a metric called CPI, the ratio of hours budgeted to a work order/actual hours taken for said work order, as the y value on the control chart with work order # as the x axis. Logically, one can determine the center line here to be 1 (if budget = actual); however I have been having problems with the upper and lower control limits.
As I said, the data is very inconsistent, and there is a tremendous amount of variability in the process at this time. So when I used a moving range to determine the control limits I got control limits somewhere in the area of LCL= 0.32 and UCL = 1.68, which are not viewed as acceptable CPI values by management, nor should they be.
With all that said, how do you determine a useful LCL and UCL when the data is highly variable?
I have been instructed to use a metric called CPI, the ratio of hours budgeted to a work order/actual hours taken for said work order, as the y value on the control chart with work order # as the x axis. Logically, one can determine the center line here to be 1 (if budget = actual); however I have been having problems with the upper and lower control limits.
As I said, the data is very inconsistent, and there is a tremendous amount of variability in the process at this time. So when I used a moving range to determine the control limits I got control limits somewhere in the area of LCL= 0.32 and UCL = 1.68, which are not viewed as acceptable CPI values by management, nor should they be.
With all that said, how do you determine a useful LCL and UCL when the data is highly variable?