Separate control limits from specifications

Separate control limits from specifications

Control limits describe expected process behaviour under a specified baseline model. Specifications express an engineering or customer requirement. They answer different questions and must be labelled differently. A process can behave consistently yet produce unacceptable output because its ordinary variation is too wide or its centre is misplaced. Conversely, a point can be within specification and still signal an unusual process change. Investigate that signal through the approved response instead of accepting it simply because the product passed. Improve the system for common-cause loss; do not continually adjust settings to chase ordinary noise. Stability must be assessed from suitable evidence, not a handful of decorative points.

Follow the method

1. Stable baseline can miss specification
2. Width (mm)
3. Sample in time order
4. Center: 10
5. LCL: 9.7
6. UCL: 10.3
7. A signal can remain within specification
8. LCL: 9.9
9. UCL: 10.1

Record product status and process signal separately

Fictional training case: quality technician Sam reviews width measurements from two different processes. The baseline center and statistical limits are supplied and approved for this exercise; the six displayed values are recent observations, not the data used to calculate those limits. Engineering specifications describe each product requirement.

Follow the method

1. A sample 1:9.82 mm
2. A sample 4:10.16 mm
3. B sample 5:10.20 mm
4. All six points

A passing part can still carry a process warning

Classify product conformance and a process signal separately, record the appropriate response, and explain what the supplied chart cannot establish.

Fictional training case: quality technician Sam reviews width measurements from two different processes. The baseline center and statistical limits are supplied and approved for this exercise; the six displayed values are recent observations, not the data used to calculate those limits. Engineering specifications describe each product requirement.

Role: Quality technician working with the process owner under a defined reaction plan.

Normal condition

The measurement, units, relevant process and chart baseline are identified before interpretation. Product acceptance is assessed against specifications; process behavior is assessed against the chart rules.

The gap

A colleague labels every in-specification observation green and proposes ignoring process B sample 5 because it still meets the drawing.

• Only the beyond-limit signal rule is illustrated; other approved pattern rules are not evaluated here.
• Baseline suitability and measurement adequacy are supplied teaching assumptions, not proven by these six points.

1. Identify two different authorities

   Sam labels the statistical limits as the supplied process baseline and the specifications as the fictional engineering acceptance range. He keeps units and process identity on both charts.

   Why: Calling both sets tolerances hides the decision being made. The source of a line determines what crossing it means.

   Evidence: Process A has narrower specifications than statistical limits; process B has wider specifications.

2. Make the product decision explicitly

   For A, he marks sample 1 at 9.82 and sample 4 at 10.16 as outside 9.90–10.10. For B, every displayed individual observation lies within 9.50–10.50.

   Why: This is a conformance comparison for the displayed values, not blanket release of all output or a claim that a process is capable.

   Evidence: A has two displayed nonconforming observations; B has none by the stated width limits.

3. Evaluate the process rule separately

   He compares each point with its own process control limits. A has no beyond-limit point. B sample 5 at 10.20 exceeds its 10.10 upper control limit by 0.10 mm.

   Why: A process signal can occur while the measured part meets specification. Conversely, a statistically predictable process can produce unacceptable output.

   Evidence: Record B sample 5 as a signal; do not redraw the baseline to include it.

4. Route the two problems

   Sam routes A nonconforming output through disposition and asks the process owner to address the supplied stable but unsuitable baseline. For B he initiates the approved signal investigation and exposure assessment.

   Why: Repeated adjustment after every ordinary fluctuation is different from investigating a signal. Neither route is replaced by a green product label.

   Evidence: Two records: product disposition and process response, each with an owner and evidence requirement.

5. Keep the inference bounded

   He notes that six displayed points cannot establish stability or estimate capability reliably. The example does not calculate control limits or capability indices.

   Why: Correct classification should not become an unsupported statement about long-term performance.

   Evidence: The final decision cites supplied baseline assumptions and asks for suitable history if those assumptions are uncertain.

The baseline belongs to a different setup

Sam learns that process B changed tooling before sample 1, and the owner has not confirmed whether the supplied baseline remains applicable.

Keep the observation and prior signal comparison visible, but mark interpretation provisional and escalate baseline suitability. Continue the applicable product controls; do not create new limits from six points.

A control limit is meaningful only for the process and statistic it represents. A baseline problem is not resolved by substituting specification limits.

Tool-change time, approved chart definition, measurement record and suitable process history are required.

Run a practice session

Materials

• Two labelled native charts and the supplied values
• Blank dual-decision record
• Ruler or highlighter; separate answer

1. Name the lines · 5 minutes

   Who or what established each line?

2. Demonstrate the two decisions · 8 minutes

   What does B sample 5 require despite meeting specification?

3. Complete changed-case record · 9 minutes

   How does observation 2 differ from 3?

4. Challenge the baseline · 6 minutes

   What if these limits came from another setup?

Debrief

• If a learner writes all good, ask whether they answered product or process status.
• Ask for the missing evidence before accepting a capability claim.

Complete the two classifications in separate columns before writing any response; then compare each with the answer.

Transfer into the work

Owner: Process owner with quality engineering

Record: Chart event and product-disposition references

Review: At the signal response and subsequent approved verification

Evidence: Correct chart identity, baseline, measurement context, affected-output assessment and disposition authority

Retain the signal/hold records and escalate the uncertainty; obtain suitable evidence before revising limits or declaring performance.