Reference

Glossary

Definitions as this course uses them. Where industry practice varies, the entry says so.

Terms as this course uses them, with the module that develops each one. Where practice varies, the entry says so rather than picking a side.

Assignable cause: See special cause.
Attribute data: Data that classifies or counts rather than measures: pass or fail, number of defects on a panel. Charted with p, np, c or u charts. Module 17.
Bias: The difference between the average of repeated measurements of a part and that part's reference value. A gauge can be precise and biased at the same time. Module 4.
Capability index: A single number comparing the spread and location of a process with a specification: Cp, Cpk, Pp, Ppk, Cpm. Meaningless without a statement of which sigma estimate produced it and whether the process was stable. Modules 7 and 8.
Common cause: Variation that is inherent in the process as it is currently run, present in every result, and not traceable to a specific event. Reducing it means changing the process, not reacting to individual points. Modules 2 and 16.
Confidence interval: A range computed from data by a procedure that, repeated on many samples, would contain the true value the stated proportion of the time. It is a statement about the procedure, not a probability attached to the particular interval you have. Module 11.
Control chart: A time-ordered plot of a statistic with limits computed from the process's own variation, used to distinguish common from special cause. Its limits are never specification limits. Modules 16 and 17.
Control limits: Lines on a control chart at the centre ± three standard deviations of the plotted statistic, computed from the data. They describe what the process does. Module 16.
Control plan: The document that says, for each characteristic, what is measured, with what, how often, by whom, recorded where, and what happens when it goes out of control. Module 18.
Cp: The specification width divided by six sigma: how well the process spread could fit the tolerance if it were perfectly centred. Says nothing about centring. Module 7.
Cpk: The distance from the mean to the nearer specification limit divided by three sigma. Accounts for centring; equals Cp only when the process is exactly centred. Module 7.
Cpm: A capability index that penalises distance from a target rather than from the nearer specification limit. Module 7.
CTQ: Critical to quality: a customer requirement translated into something specific and measurable on the process. Module 3.
Defect: A single nonconformity. One unit can carry several. Counted with c and u charts. Module 6.
Defective: A unit that has at least one defect. Counted with p and np charts. Module 6.
DMAIC: Define, Measure, Analyze, Improve, Control: the improvement cycle this course is organised around. Module 0.
DPMO: Defects per million opportunities. Depends entirely on how opportunities are counted, which is why it is easy to game. Module 6.
DPU: Defects per unit. Module 6.
EWMA chart: Exponentially weighted moving average chart: each point carries a weighted memory of earlier points, which makes it far quicker than a Shewhart chart at finding a small sustained shift. Module 17.
FMEA: Failure mode and effects analysis: a structured review of how a process or product can fail, how bad it would be, how often it happens, and how likely it is to be caught. Module 10.
Gauge R&R: A study that splits measurement variation into repeatability (the same operator, the same part) and reproducibility (different operators), and compares it with the part variation and with the tolerance. Module 4.
Individuals chart (I-MR): A control chart of single measurements with a moving range chart beneath it, used where subgrouping is impossible or meaningless. Module 16.
Kappa: A chance-corrected measure of agreement between raters on a categorical judgement, used in attribute agreement analysis. Module 4.
Lean: A body of practice aimed at removing waste and shortening flow time. Complementary to Six Sigma's focus on variation, and not the same thing. Modules 0, 2 and 14.
Linearity (of a gauge): How the gauge's bias changes across its working range. Module 4.
Multi-vari chart: A plot that separates positional, cyclical and temporal variation so you can see which family dominates before testing anything. Module 9.
ndc: Number of distinct categories: how many genuinely different levels a gauge can distinguish across the part variation. Five or more is the usual guideline. Module 4.
Operational definition: A statement of how a characteristic is measured and classified, specific enough that two people applying it to the same part reach the same answer. Module 5.
p-value: The probability of data at least as extreme as what you observed, if the null hypothesis were true. Not the probability that the hypothesis is true, and not a measure of effect size. Module 11.
Poka-yoke: A design or fixture feature that makes an error physically impossible or immediately obvious, rather than relying on care. Module 14.
Pp, Ppk: Capability indices computed from the overall sample standard deviation. Under the current AIAG & VDA and ISO 22514 convention, the letter also records that stability has not been demonstrated. Modules 7 and 8.
PPAP: Production part approval process: the automotive submission package that includes initial capability studies, MSA results and a control plan. Module 18.
Rational subgrouping: Choosing subgroups so that within-subgroup variation contains only the routine variation you want the limits to represent, and anything you want to detect appears between subgroups. It decides what the chart can and cannot see. Module 5.
Repeatability: Variation when the same operator measures the same part with the same gauge more than once. Module 4.
Reproducibility: Variation between operators measuring the same parts with the same gauge. Module 4.
RTY: Rolled throughput yield: the product of the first-time yields of every step, so it counts rework the final yield hides. Module 6.
Run rules: Patterns other than a point outside the limits that also signal a special cause, such as eight points in a row on one side of the centre line. Each rule added raises sensitivity and the false alarm rate together. Module 16.
Sigma level: A defect rate expressed as a Z value. Quoted both with and without the 1.5 shift, which is why the same process is described as four sigma or five and a half by different people. Module 6.
SIPOC: Supplier, input, process, output, customer: a one-page scope of a process, used early in Define. Module 2.
Special cause: Variation traceable to a specific, identifiable event or change, which a control chart is designed to detect. Modules 2 and 16.
Specification limits: The boundaries of what the customer or the design will accept, applied to individual parts. They come from the drawing, never from the data, and never appear on a control chart. Modules 7, 16 and 18.
Stability (of a gauge): The condition of a measurement system giving the same answer on the same reference part over time. Checked with a control chart of repeated readings. Module 4.
Stability (of a process): The condition of showing only common cause variation over the period observed, so that one distribution describes the process and predictions from it mean something. Required before any capability index is interpreted. Modules 7, 8 and 16.
Standard work: The documented, current best-known method for a task, kept close enough to reality that operators actually follow it. Module 14.
Taguchi loss function: A quadratic model of the cost of being away from target, used to argue that loss starts as soon as a part is off target, not when it crosses a specification limit. Module 15.
Tolerance: The total permitted range of a characteristic: USL − LSL. Module 7.
VOC: Voice of the customer: what customers actually say they need, before it is translated into measurable requirements. Module 3.
Western Electric rules: The original set of supplementary control chart signals from the Western Electric statistical quality control handbook, later extended by Nelson. Module 16.
Within-subgroup sigma: An estimate of the process standard deviation built only from variation inside subgroups, usually R̄/d₂ or S̄/c₄. It excludes drift between subgroups, so it is generally smaller than the overall sigma. Modules 5, 7 and 16.
Xbar-R chart: A control chart of subgroup means with a range chart beneath it, the standard variables chart for subgroups of about two to nine. Module 16.