Locking Control Limits and SPC Sample Sizes (AIAG/Automotive Requirements)


Starting to get Involved
Does anyone know of automotive OEM requirements on when to lock control limits and/or requirements of the number of samples required prior to calculating control limits?

We have been discussing the phenomenon of a stable process with control limits narrowing to the point of a nonconformance occurring due to a point being out of control. I've seen reference to "locking" control limits and am wondering if anyone is familiar with either IATF/AIAG or automotive OEM requirements on how/when this is done.

Thank you,


Steve Prevette

Deming Disciple
Super Moderator
I would definitely NOT recommend "locking" the control limits. They would then simply become specification limits and would not be based upon the current data. If the variation has been reducing with time due to improvements (yea!) you should follow the data and narrow the control limits if you understand what the change was that cause the variation to reduce. That is an important point Dr. Wheeler pushes - you do want to be sure you know WHY the process changed before you change the control limits. In my experience, even if you do not know, you eventually need to "follow the data".

If a process has been improving, with improved data and control limits in your SPC trending, you would still want to examine why you got an out of control (but perhaps in specification) result. Graded approach is acceptable - but the idea still is to head off a problem BEFORE you end up making out of specification components.


If you do not lock the control limits you run the risk of missing important signals. If you have a process that is slowly trending in one direction and you do not have your limits locked, the limits will continue to shift with the process and you will have what would have been out of control points that will not be tagged as out of control. Same thing if you have a process where the variation is gradually growing. It's like the old story about the frog not realizing it's boiling in the pot of water that slowly heats up. I've run into this problem using a certain MRP program with a built in control chart feature that keeps recalculating the control limits.
Top Bottom