PSLE-SCI-REALITY-0121
Wait, What? A Process Can Be Perfectly Stable—and Reliably Make the Wrong Thing
A fictional factory makes small rods that are supposed to be between 9.8 mm and 10.2 mm thick.
For an entire morning, the machine produces rods near 10.35 mm. The measurements are remarkably steady:
| Rod | Measured thickness |
|---|---|
| 1 | 10.34 mm |
| 2 | 10.36 mm |
| 3 | 10.35 mm |
| 4 | 10.33 mm |
| 5 | 10.36 mm |
| 6 | 10.35 mm |
A dashboard gives the machine a green badge:
PROCESS IN CONTROL
Can the factory now say every rod meets the specification?
No.
The process can be stable around the wrong centre. “In control” is mainly about whether the process is behaving consistently relative to its own established pattern. “Meets specification” is about whether the product falls inside the required acceptable range. These are related quality questions, but they are not the same question.
The Reality Lab habit is: when you see a quality-control badge, identify which limits belong to the process and which limits belong to the product requirement.
Quick Answer
- Find the specification limits: what values count as acceptable for the product or outcome?
- Find the control limits: what range is being used to judge whether the process is behaving like its established pattern?
- Do not assume the two sets of limits are identical or come from the same source.
- Check the centre of the process as well as its spread. A narrow, stable process can still be centred outside the acceptable product range.
- Check individual product results. A process can be statistically stable while some or many units fail specifications.
- Keep the conclusion precise: “the process appears stable” is not the same as “every product is acceptable”.
The Exact Learner Job This Page Owns
This Reality Lab page owns one real-world communication object: a quality dashboard, factory chart or product claim uses “in control” as if it means “all products meet requirements”.
It does not become a statistical-process-control textbook. The broader mathematical owner remains Real-World Mathematics: Manufacturing Tolerances, Yield and Quality Control, and the wider threshold concept is explained at How Thresholds Work.
For Primary 5/6 learners, the job is evidence transfer: read the claim object, distinguish the two kinds of boundary, and stop the green badge from replacing the underlying science.
- Reality Lab Vol No.067: “Within Tolerance” — Does Passing Mean Exactly on Target?
- Reality Lab Vol No.080: The Average Passed — Did Every Item Pass?
- Reality Lab Vol No.107: “QA/QC Passed” — Does That Mean Every Result Is Correct?
- How to Tell a PSLE Science Method Limitation From a Mistake
Original Reality Lab Case: The Steady Machine That Missed the Target
This is an original composite teaching case with constructed data. It does not reproduce a real factory chart or proprietary quality system.
A fictional machine cuts plastic strips. The product requirement says each strip should be between 49.5 mm and 50.5 mm long.
During a stable operating period, the machine repeatedly produces strips around 51.0 mm:
| Strip | Length | Meets product specification? |
|---|---|---|
| A | 50.96 mm | No |
| B | 51.03 mm | No |
| C | 50.99 mm | No |
| D | 51.01 mm | No |
| E | 50.97 mm | No |
| F | 51.04 mm | No |
The readings are close together. The process may therefore be highly consistent. But it is consistently centred too high for the product requirement.
This is the core Reality Lab surprise: low variation is not enough if the stable centre is wrong.
Observed, Claimed and Inferred
| Layer | Statement |
|---|---|
| Observed | The recent measurements form a stable-looking cluster near 51.0 mm. |
| Process interpretation | No unusual shift beyond the process-monitoring rule is currently evident. |
| Product requirement | Acceptable length is 49.5–50.5 mm. |
| Unsupported claim | “The process is in control, therefore every product meets specification.” |
| Better conclusion | The process may be stable yet centred outside the acceptable product range, so stability and conformity must be checked separately. |
Two Different Questions Need Two Different Boundaries
A specification limit answers a question such as:
What product values are acceptable for the intended requirement?
A control limit answers a different question such as:
Is the process behaving within the range expected from its established pattern of variation?
NIST’s statistical process-control guidance makes this distinction clearly. Control limits come from the behaviour of the process or a statistical model of that behaviour. Specification limits come from product, engineering, customer or other acceptance requirements.
The two can sometimes be drawn on the same chart. That does not make them the same kind of line.
The Source Check: Who Created Each Limit?
Whenever a scientific or technical graphic shows a limit, ask where the line came from.
| Limit | Typical source of the boundary | Main job |
|---|---|---|
| Specification limit | Design requirement, agreed tolerance, product requirement or stated acceptance rule | Decide whether output is acceptable for the requirement |
| Control limit | Observed process variation and a statistical monitoring rule | Detect unusual changes in process behaviour |
A control limit is not automatically a promise that everything inside it is good. A specification limit is not automatically evidence that the process producing the item is stable.
The Stable-but-Wrong Case
Imagine a machine that should fill bottles with 500 mL. It steadily fills them with 480 mL, with very little variation.
From one viewpoint, the machine is consistent. From another, the output is unacceptable.
A process-control chart may show no unusual change because 480 mL is its current stable pattern. The product specification can still reject the bottles.
Consistency and correctness are different properties—just as precision and accuracy are different measurement ideas.
The In-Spec-but-Unstable Case
The opposite problem can also occur.
Suppose the acceptable product range is 9.0–11.0 units. A process usually runs close to 10.0, but its recent sequence suddenly shifts from around 9.8 to around 10.8. Every individual product may still be inside specification for the moment.
A process-monitoring system may nevertheless flag the shift because the behaviour is unusual compared with the previous process pattern. That warning can be useful before products actually cross the specification boundary.
So “this item passes specification” does not prove “the process is stable”, just as “the process is stable” does not prove “this item passes specification”.
The Representation Check: Four Lines Can Look Like One Kind of Boundary
A quality chart can show an upper control limit, lower control limit, upper specification limit and lower specification limit. If they use similar colours or labels, a young reader may see four horizontal lines and assume all four mean “pass/fail”.
Before interpreting the graph, label each line by its scientific job.
- Which lines describe expected process behaviour?
- Which lines describe acceptable product output?
- Which values are actual measurements?
- Is the process centre inside the specification range?
- Is the process spread narrow enough relative to the specification range?
The visual position of the lines matters only after their meanings are known.
The Baseline Check: “In Control” Depends on a Reference Pattern
Calling a process “in control” requires some idea of what stable behaviour looks like for that process. The monitoring system is therefore comparing new data against a reference pattern or model.
If the historical process was already centred incorrectly, a control chart can faithfully show that the machine is continuing to behave like its old, wrong-centred self.
This is why process history cannot replace product requirements.
The Average Check: A Stable Mean Can Hide Individual Failures
A dashboard may plot the average of several products rather than every individual unit. The average can remain inside specification while some individual products lie outside it. Vol No.080 owns that group-average problem.
The lesson for this article is that even after process-control status is understood, the learner must still ask what the plotted point represents: one unit, a group average, a range, or another summary.
The Measurement Check: Can We Trust the Product Value Near a Boundary?
If a product measures 10.20 mm and the upper specification limit is 10.20 mm, measurement uncertainty can matter to the pass/fail decision. Reality Lab Vol No.067 already owns the idea that passing a tolerance does not mean the item is exactly on target.
This article keeps that as a routed micro-skill rather than re-teaching conformity assessment. The key point is simply that process stability, product specification and measurement confidence are separate layers.
Alternative Explanation 1: The Process Has Shifted but the Products Still Pass—for Now
An unusual shift can be an early warning. If the process is moving toward a specification boundary, current products can remain acceptable while future failure risk increases.
Alternative Explanation 2: The Process Is Stable but the Machine Was Set Incorrectly
A stable machine can be adjusted to the wrong target. Its measurements then remain consistent yet off-centre. This is the classic “stable but unacceptable” case.
Alternative Explanation 3: The Process Is Stable but Too Variable
A process can have a stable centre near the target but a spread so wide that many individual products fall beyond specification limits. Stability does not automatically mean the variation is small enough for the required tolerance.
Alternative Explanation 4: The Quality Badge Summarises Only One Part of the System
A green “process in control” badge may refer only to the monitored process statistic. Product inspection, measurement uncertainty, specification conformity, material quality and final-use performance may be evaluated elsewhere.
One badge can therefore be correct while a much broader advertising sentence built from it is wrong.
What Evidence Would Strengthen a Claim That the Process Produces Acceptable Products?
- The specification limits and their source are stated.
- The control limits and monitoring rule are stated separately.
- The process centre is shown relative to the specification range.
- The spread of individual output is shown or otherwise evaluated.
- Individual product conformity is checked where the requirement concerns individual units.
- Measurement capability is suitable for the tolerance being judged.
- Recent process behaviour is compared with a justified stable baseline.
- Out-of-control signals are investigated rather than hidden.
- A green process-status badge is not used as a substitute for product-performance evidence outside the measured quality characteristic.
What Would Weaken It?
- The chart labels control limits as though they were product requirements.
- The specification range is never shown.
- The process centre lies outside specification but the report focuses only on stability.
- Group averages are shown while individual failures are hidden.
- The monitoring baseline was built from a process known to be incorrectly centred.
- A process change is dismissed because current items still happen to pass.
- The quality badge is used to make unrelated claims about safety, durability or effectiveness.
Worked Case 1: Stable but Too High
A fictional sensor component should measure 20.0 ± 0.5 units. A machine produces 20.8, 20.7, 20.8, 20.9, 20.8 and 20.7. The results are tightly grouped. The process may be stable, but every listed component lies above the upper product specification of 20.5. Stability cannot rescue conformity.
Worked Case 2: In Specification but Process Shifted
A product specification is 95–105 units. A process previously centred near 100 suddenly produces a long run near 104. All current units may still pass. The shift is nevertheless scientifically meaningful because the process is behaving differently and is much closer to the boundary than before.
Worked Case 3: Same Average, Different Risk
Process A produces 99.9, 100.0 and 100.1. Process B produces 95, 100 and 105. Both can have the same average. If acceptable output is 98–102, the averages hide radically different product conformity. The spread matters.
Worked Case 4: The Badge That Travels Too Far
A factory dashboard says its fill-volume process is “in statistical control”. An advertisement changes this into “scientifically proven superior durability”. That claim has jumped evidence jobs. Stable fill volume provides no direct evidence about how long the product lasts.
Tempting Reasoning That Fails
- “Green means every product passed.” The green status may refer only to process stability.
- “Inside control limits means inside specification.” The two sets of limits can have different positions and meanings.
- “Stable means accurate.” A process can be stably centred at the wrong value.
- “Every current item is in specification, so the process must be stable.” A process can shift while remaining temporarily inside broad product limits.
- “Control limits are chosen by the customer.” They generally arise from process behaviour and the monitoring model; specifications arise from requirements.
- “Specification limits prove the process can meet them.” A requirement states what is wanted, not whether the process is capable of reliably producing it.
Model and Measurement Limits
Formal statistical process control uses specific chart types, assumptions and rules. Those details belong to specialist Mathematics and quality-engineering owners. Reality Lab deliberately keeps only the transferable evidence distinction.
Control limits are not universal fixed percentages. Their calculation depends on the chart, process data and modelling assumptions. Specification limits are also context-specific. A Primary learner should therefore avoid memorising invented numeric rules and focus instead on the source and purpose of each boundary.
How Far Can the Conclusion Travel?
If a process-control chart shows no evidence of unusual process behaviour under its monitoring rule, a cautious statement is: the monitored process appears statistically stable relative to its established pattern.
To conclude that products meet specification, we must separately compare product output with the actual acceptance limits. To conclude that every unit meets specification, we need evidence at the unit level or a justified system capable of supporting that claim. To conclude that the product is safe, durable or effective, we need evidence about those different outcomes.
PSLE-Style Transfer Case
A machine makes rods. The acceptable length is 19.5–20.5 cm. During one hour the machine repeatedly produces rods between 20.7 and 20.9 cm with very little variation. A pupil says, “The process is very consistent, so the rods must be acceptable.”
Question: Explain why the pupil’s conclusion is wrong.
Reasoned answer: Consistency describes how similar the produced lengths are to one another. Acceptability depends on whether the lengths fall within the required specification of 19.5–20.5 cm. The rods are consistently above the upper specification limit, so a stable pattern does not make them acceptable.
Explained Practice
Practice A: A process is stable around 8.0 units, but the required product range is 9.5–10.5. What is the main problem? The process is centred outside the specification range.
Practice B: All products currently fall inside a broad specification, but measurements suddenly shift upward compared with the historical process. Should the shift be ignored? No. Process stability is a separate question and an unusual shift can be an early warning.
Practice C: A dashboard shows control limits but no product requirements. Can a learner conclude all output is acceptable? No. The specification boundary is missing.
Delayed Independent Return: The Two-Limits Test
Tomorrow, when you see a quality graph, do not begin with the coloured badge. Begin with two questions:
- What limits describe the process? These help tell us whether behaviour has changed unexpectedly.
- What limits describe acceptable output? These tell us whether the product meets the stated requirement.
If the graph cannot answer both, do not let one boundary silently pretend to be the other.
Parent and Tutor Teaching Guide
Draw two target boxes on paper. In the first, make a very tight cluster of six dots that sits completely outside the target rectangle. Ask, “Are the dots consistent?” Yes. “Are they on target?” No.
In the second, place scattered dots across the whole target rectangle. Ask, “Are all dots currently inside the acceptable area?” Perhaps yes. “Is the pattern as stable and predictable as the first cluster?” Not necessarily.
This visual activity builds the distinction without formulas. Then introduce the real-world language: control limits describe process behaviour; specification limits describe acceptance requirements.
For an advanced learner, add a third layer: measurement uncertainty near the boundary. Ask why even a product specification decision can require another evidence check when the measured value lies extremely close to the limit.
Authoritative Sources
- Singapore Examinations and Assessment Board — 2026 PSLE Science Syllabus
- Ministry of Education, Singapore — 2023 Primary Science Teaching and Learning Syllabus
- NIST/SEMATECH e-Handbook — Control Chart Principles
- NIST/SEMATECH e-Handbook — Process Capability and Specification Limits
NIST distinguishes statistical control from process capability relative to specification requirements. For a Primary Science learner, the deeper habit is familiar: identify what each line or label represents, trace it back to the evidence job it was designed to perform, and never let a convenient summary badge support a broader claim than the underlying method can justify.
The Quiet Return
A machine can repeat itself beautifully and still miss the target.
A product can meet today’s target while the process has started behaving strangely.
In science, “stable” and “acceptable” are different claims. Check the boundary that belongs to each one.
