Engineering & Technology · Quality Engineering, Experiments & Process Evidence
Variation, specifications and definitions of quality
Two groups of parts have the same average length. One group fits a supplied interval; the other includes pieces that are too short and too long. The mean has hidden a consequential difference. Quality reasoning begins with the service and the individual characteristics needed to provide it.
Review measurement and uncertainty first. Means, ranges, square roots and percentages are used directly. The inference lesson defines its paired-difference model and supplies its Student-t quantile; mathematics develops probability and inference further. These examples teach interpretation rather than prescribe production sampling, monitoring or release rules.
Back to the chapterAfter reading, you can
Define a quality characteristic and distinguish individual variation, averages and specification decisions.
Name the characteristic and its requirement
A quality characteristic is an attribute being evaluated, such as length, resistance, response time, readability or absence of a defined defect. State the item, quantity, measurement conditions and requirement. A dimension may be critical to one function while appearance or accessible operation matters to another. Passing one selected characteristic cannot establish overall suitability.
A specification gives required limits or conditions. It should have a justified relationship to function, compatibility or another obligation. A measured process average does not set those limits by itself. If the process produces inconvenient results, simply widening the requirement changes the obligation rather than demonstrating an improved product.
Compare more than the mean
Our original observations are A: 49, 50, 51 mm and B: 48, 50, 52 mm. Both means are 50 mm. Assume a supplied illustrative interval from 49 through 51 mm inclusive and a decision using point observations only. All three A values and only the middle B value fall inside that interval. The second group has a greater range, 4 mm rather than 2 mm.
These are tiny constructed groups, not evidence of stable production or a defect-rate estimate. Their purpose is to show why location and spread answer different questions. The figure preserves individual values and labels its specification boundaries explicitly. Measurement uncertainty and an actual conformity decision rule would require further attention in a real acceptance decision.
Variation needs an account
Variation can arise between parts, within an item, across operating periods and through the measurement process. Retain identifiers and the conditions needed to distinguish those sources. Mixing startup, routine operation and a different material batch into one unlabeled histogram can erase the structure that explains the observations.
A process can be statistically consistent yet consistently fail its requirement. Conversely, a few observed items inside limits do not establish that the process will remain stable. Preserve time order and examine relevant distributions and changes before extending observations into a prediction. Describe the sample and evidence limitations rather than applying a reassuring quality label to a single average.
Keep three questions separate
The constructed sums are 49+50+51=150 and 48+50+52=150, each divided by 3 giving 50 mm. The ranges are 51−49=2 mm and 52−48=4 mm. The point-value interval count is 3 of 3 for A and 1 of 3 for B under the explicitly supplied inclusive rule.
These calculations answer location, spread and observed point-value conformity. They do not answer whether true dimensions comply once uncertainty is considered, whether future items will comply or whether the complete product performs its intended service. A useful report names the question each number answers and leaves the unsupported conclusions open.
Try the reasoning
Does an average inside the specification establish that every observed item is inside?
Show the answer and reasoning
No. The original B group has mean 50 mm while two individual values lie outside the supplied interval.
Can a stable process still be unsuitable?
Show the answer and reasoning
Yes. Stable variation can be centred incorrectly or be too wide for the requirement.
Why retain the measurement conditions with part observations?
Show the answer and reasoning
They help distinguish product variation from measurement effects and determine which comparisons are meaningful.
What to carry forward
Define the required service and characteristic, then preserve the individual, temporal and measurement information needed to interpret variation.
References
Teaching resources and research supporting this lesson’s explanations and examples.
National Institute of Standards and Technology (NIST)3 sources