© 2004 Cayman Business Systems
Rev: Pre-G3 - Rendered  Thursday, February 12, 2004
Slide 47
Elsmar.com  --- The Cove!
8-Disciplines Problem Solving
Elsmar.com
•Tampering is additional variation caused by unnecessary adjustments made in an attempt to compensate for common cause variation.
•Tampering with a process occurs when we respond to variation In the process (such as by “adjusting” the process) when the process has not shifted. In other words, it is when we treat variation due to common causes as variation due to special causes. This is also called “responding to a false alarm,” since a false alarm is when we think that the process has shifted when it really hasn’t.
•In practice, tampering generally occurs when we attempt to control the process to limits that are within the natural control limits defined by common cause variation. We try to control the process to specifications, or goals. These limits are defined externally to the process, rather than being based on the statistics of the process.
Tampering - Process Variation
Deming showed how tampering actually increases variation. It can easily be seen that when we react to these false alarms, we take action on the process by shifting its location. Over time, this results in process output that varies much more than if the process had just been left alone.

Rather than using the suggested control limits defined at ±3 standard deviations from the center line, we instead choose to use limits that are tighter (or narrower) than these (sometimes called Warning Limits). We might do this based on the faulty notion that this will improve the performance of the chart, since it is more likely that subgroups will plot outside of these limits. For example, using limits defined at ±2 standard deviations from the center line would produce narrower control limits than the ±3 standard deviation limits. However, you can use probability theory to show  that the chance of being outside of a ±3 standard deviation control limit for a Normally-distributed statistic is 0.27% if the process has not shifted. On average, you would see a false alarm associated with these limits once every 370 subgroups (=1/0.0027). Using ±2 standard deviation control limits, the chance of being outside the limits when the process has not shifted is 4.6%, corresponding to false alarms every 22 subgroups!