Tampering on the Shop Floor: How a Control Chart Raised Scrap
Precision Parts machines a bore diameter with a CTQ of 25.00 ±0.05 mm, at about 300 parts per shift. An X-bar and R chart was introduced with subgroups of five every 30 minutes. Operators were trained to 'keep the chart in the green' and to adjust the machine offset whenever a plotted subgroup mean falls outside one standard deviation of the centreline. Over ten weeks scrap rose from 1.1% to 2.6%, even though incoming material and the machine itself are unchanged. Process capability before the chart was introduced was Cp 1.4. The control plan states only 'operator adjusts as needed' and contains no reaction plan. The shift supervisor now wants to add a second chart for the range and to tighten the specification limits so that fewer parts fall outside. Audit records show that 80% of adjustments were made on points well inside the control limits.
1. Diagnose what has happened to the process. Use the language of common and special cause variation.
Read the worked answer
This is over-adjustment, or tampering. The chart has control limits at three sigma, but operators were instructed to react at one sigma - so they respond to routine, common-cause noise as if it were a signal. Every unnecessary adjustment adds a new disturbance to the process, and those disturbances accumulate; the variation of the output rises even though no input changed. The evidence fits precisely: 80% of adjustments were triggered by points inside the control limits, capability was already 1.4 before the chart, and scrap tripled without any change in machine or material. The principle at stake is that common-cause variation is owned by the system and must not be corrected point by point at the machine; only special causes - points beyond the control limits, runs, trends - justify action. Tightening specs would not help: specification limits are the customer's requirement, not a process control device, and shrinking them just relabels good parts as scrap.
2. What should the reaction plan say, and who owns each part of it?
Read the worked answer
The reaction plan should define, in writing: (1) the signal - a point outside the three-sigma control limits, a run of seven or more on one side of the centreline, a trend of seven, or a pattern rule the team has chosen; (2) the immediate response - stop and contain only if the signal indicates a real special cause, notify the shift leader, record the time, the subgroup and the suspected cause; (3) the investigation - check the documented list of likely special causes (setup, tool wear, material lot, gauge, operator change); (4) the authority - who may change the offset and under what conditions; (5) the escalation - when the quality engineer or maintenance is called; (6) the record - what goes into the log and how it feeds the next shift. Ownership: production owns detection and containment, quality engineering owns the rules and their revision, maintenance owns equipment causes, and the process owner signs off changes. Crucially, the plan must state that points inside the control limits are not adjusted.
3. The supervisor wants a tighter specification and a second chart. How do you respond, and how would you verify that the fix has worked?
Read the worked answer
Decline both as stated, with reasons. Specification limits come from design and customer requirements; changing them to reduce apparent scrap conceals the problem and puts nonconforming parts into the field. A range chart is already part of the X-bar and R pair - if the R chart is not being plotted and interpreted, the fix is to use it properly, since a rising range is exactly the signature of tampering. The corrective actions are: retract the one-sigma rule, retrain operators and supervisors on the actual signal rules, revise the control plan and reaction plan, and remove or lock the offset adjustment so changes require authorisation. Verify by re-baselining: track scrap and within-subgroup standard deviation for at least four to six weeks after the rule change, confirm the range chart is stable and capability returns to at least the earlier Cp 1.4, then recalculate control limits from the stable period. Only then decide whether the process needs a genuine capability improvement project.