Garment and Textile Quality Inspection Assessment for Production, Quality Control and Sourcing TeamsSixteen garment defects. Minor, major or critical — and whether your severity calls move with the standard.
A 15 mm thread inside a side seam. A 20 mm run of broken stitching on a back rise. A bead that comes off an infant's sleepsuit under a light pull. A colour reading inside the tolerance. Each one described in words, each one answered with the same four calls, so the standard is on the screen rather than in your memory. Twenty-four more exercises on what a sampling result actually permits, how a measurement sits against its tolerance, what a repeated defect points to and what to do with the lot. No brand, buyer, retailer, mill or machine is named. Reported as a rack of confidence ribbons, where every figure is a band and two bands that overlap are printed as not distinguishable rather than ranked.
A garment quality inspection test that scores whether your severity calls match the standard, not whether you can recite a defect list
The Garment and Textile Quality Inspection Assessment is a thirty-minute check of whether an inspector's severity calls match the standard: sixteen described defects called as nothing, minor, major or critical, plus twenty-four exercises on what a sampling result permits, how a measurement sits against its tolerance, and what to do with the lot.
Nobody on an inspection floor argues about whether the thread is there. They argue about what it is worth. So this instrument does not ask you to spot a defect; it describes one completely — the garment, the place, the size, whether the seam still holds — and asks the only question that actually costs money: which rung does this sit on? The same four calls appear on every one of the sixteen, with the test each rung has to meet written into the option, so what is being measured is your judgement and never your memory of a manual.
Your sixteen calls are read as one vector and correlated with the key. A correlation does not change when a vector is shifted or stretched, so how high up the ladder you sit and how much of it you use both drop out of the headline with nothing adjusted in your answers. That matters, because the two commonest disputes between an inspector and a supervisor are exactly those two things, and they are separate questions from whether your calls move with the standard.
Zero on the scale is not a coin. Every option, claim, pair and position in the form carries a declared share — the way working inspectors are expected to answer it — and the headline is corrected against a seeded draw from those shares, so zero means no better than the ordinary inspector and one hundred means the key exactly. A respondent who gave the same rung to all sixteen has no shape to correlate; the match is refused in words, in the place the figure would have taken, rather than printed as a zero.
Because a correlation discards level information entirely, the report prints two further figures separately and labels them descriptive rather than scored: whether you call defects generally harsher or generally softer than the key, and whether you use more or less of the ladder. The page says in plain words that the match cannot see either of them. A person who calls everything one rung harsher has the same correlation as a person who calls everything level, and an inspection floor needs to know which one it has.
Every figure is drawn as a band, never as a point: an outlined box at plus or minus 1.96 standard errors with the estimate as a faint tick inside it, a hatched box at plus or minus one, and the plain sentence that about two chances in three of a re-inspection would land inside the inner band. The bands carry an outline and no gradient, because a gradient bands on the laser printer where a quality report is actually read. Where two areas' bands overlap the page says they are not distinguishable at this length instead of ranking them.
Nothing here depends on one buyer's private manual, one brand's own standard or one country's factory law, and none of them is named. Every rule is stated inside the exercise that applies it, as the standard your team uses. The acceptance-sampling concept is public and is taught and assessed freely; no standard's tables are reproduced, and the instrument is not affiliated with any standards body, certifying body, buying house or software vendor. It is not an inspector's certification and does not qualify anybody to release a shipment.
What you walk away with
Your sixteen severity calls against the key, as a correlation corrected against an inspector answering at the declared marginals, with both the 68 and the 95 per cent spans drawn on the axis and written in words. Sixteen calls clears the fifteen a per-person correlation needs; below that the report says so and prints no figure.
Your mean rung against the key's as a signed lean, with the honest cost of each direction written out: a lot held for a fault a buyer would have accepted never appears in a defect report, and neither does a unit that got past you until it comes back. Beside it, exact agreement rung for rung, chance-corrected, and your spread against the key's. Crowding into minor and major is the commonest way an inspection record stops being a decision, and both figures are printed apart from the match because a correlation cannot see either.
A loose thread inside a seam against an open seam at an armhole; a slub in a low panel against a permanent mark on a front chest panel; a needle point in a lining and a bead that comes off an infant's garment; and a reading that sits inside its tolerance, which is the standard met rather than a fault.
Four major defects against a limit of three; a sample drawn from the front of the pallet; how a sample size moves when a lot grows tenfold; and the one sentence that separates a passed lot from a clean lot in both directions.
A chest 1 cm over plus tolerance with a sleeve inside it; a point the chart never named; the same ten trousers measured twice with 1.5 cm between the sets; and which of those differences is the garment and which is the tape.
One skipped stitch on one line's output; shade differing between a sleeve and a body panel; waist out of spec on one size only; what has to be true before a sorted lot is offered again; and why repair and cause are two separate pieces of work.
Inside your report
Illustrative sample — your report is generated from your own responses.
In words: about two chances in three that a re-inspection would land between 57 and 79 on the severity match, and about nineteen in twenty between 46 and 90. The measurement and the level bands overlap, so the report prints that they are not distinguishable at this length rather than ranking them.
| Figure | Estimate | 68% band | 95% band | Or the reason there is none |
|---|---|---|---|---|
| Severity match (shape) | 68 | 57 to 79 | 46 to 90 | — |
| Defect classification and… | 41 | 31 to 51 | 22 to 60 | — |
| Sampling and acceptance d… | 55 | 42 to 68 | 30 to 80 | — |
| Measurement and tolerance | 47 | 34 to 60 | 22 to 72 | — |
| Disposition and root cause | — | — | — | No figure: 5 of 8 answered, under the 8 this instrument needs before it prints a number. A placement is printed in its place and carries no percentile. |
Outlines and hatches rather than gradients, because a gradient bands on the laser printer where a quality report is actually read. An area under eight answered exercises carries a three-way placement and no number.
You sat 0.44 of a rung above the key on average. You are the inspector who does not let things through, and the cost of that is invisible: a lot held for a fault a buyer would have accepted never appears in any defect report, and the line waits anyway.
| Rung | The test it has to meet | Key | You | Reading |
|---|---|---|---|---|
| ○ 0. Not a defect | the reading sits inside the tolerance | 3 | 1 | ▼ used less often |
| ◔ 1. Minor | a visible departure, the unit stays usable | 5 | 4 | ▼ used less often |
| ◑ 2. Major | usability or the life of the unit is reduced | 5 | 7 | ▲ used more often |
| ● 3. Critical | the unit could hurt the person wearing it | 3 | 4 | ▲ used more often |
A correlation does not change when a vector is shifted or stretched, so the match above cannot see any of this. That is why the lean and the ladder are printed here, labelled descriptive, and never folded into a score.
Built for
- Garment quality-control inspectors and QC supervisors who want to know whether their severity calls line up with the standard rather than with the last argument they had
- Production and finishing line leads who carry the cost of a held lot and the cost of a returned one, and need both directions named
- Buying-house and sourcing quality staff who inspect against a written standard and have to defend a call to a factory and a customer in the same afternoon
- Apparel training institutes and in-house QC trainers who need a provider-neutral instrument with a printed refusal of what it did not measure
- Hiring managers screening for an inspection role, who want the severity judgement tested on its own before anything else
Find out whether your severity calls move with the standard
40 exercises across five formats · about 30 minutes · sixteen severity calls matched as a shape, two descriptive figures the match cannot see, and four areas corrected against the ordinary inspector. Every figure drawn as a band.
₹499 (incl. GST) · assessment and full report, nothing further to pay
Frequently asked questions
The four severity options are printed in a fixed ladder on every call, so the option you pick is the severity level you assigned, and your sixteen picks form one vector. That vector is correlated with the key's, and the correlation is corrected against the expected correlation of an inspector whose answer to every call is drawn from that call's declared marginal, by a seeded Monte Carlo. Zero is that ordinary inspector and one hundred is the key exactly. A respondent who gave the same rung to all sixteen has no shape to correlate and is refused a figure in words rather than given a zero.
Because a correlation does not change when a vector is shifted or stretched. An inspector who calls everything exactly one rung harsher than the key has the same correlation as one who calls everything level with it, and that is a feature rather than a flaw: it means your anchoring and your use of the scale are not being scored as if they were judgement. The report therefore prints the signed lean and the spread ratio separately, labels them descriptive rather than scored, and says in plain words that the match cannot see either of them.
No. Every rule is stated inside the exercise that applies it, as the standard your team uses, and no brand, buyer, retailer, mill, machine, software or inspection scheme is named anywhere. The acceptance-sampling concept is public and is taught and assessed freely here; no standard's tables, figures or clauses are reproduced, and the instrument is not affiliated with or endorsed by any standards body, certifying body, buying house or vendor. Nothing in it depends on one country's factory law.
Because the largest single source of complaint about any report is two people comparing a 68 with a 71. A band makes the comparison honest: each figure carries an outlined box at plus or minus 1.96 standard errors of measurement and a hatched box at plus or minus one, with the sentence that about two chances in three of a re-inspection would land inside the inner band. The bands are outlines and hatches rather than gradients, because a gradient bands on a laser printer. Where two areas' bands overlap the report prints that they are not distinguishable at this length instead of ranking them.
About thirty minutes for forty exercises across five formats. The sitting is free to take; the report is the product, priced at ₹499 in India, inclusive of GST, or US$4.99 elsewhere, one time. An unanswered exercise leaves the numerator, the denominator and the chance term together rather than scoring zero, and an empty sitting scores exactly zero because it is no evidence rather than poor judgement. A sitting with fewer than twenty-four of the forty answered is not reported at all: the refusal is printed where the ribbons would have been, and fewer than fifteen severity calls carries no match, because below fifteen a per-person correlation is unstable rather than merely imprecise.
Each one takes a single capability, puts you inside the situations where it is actually tested, and scores your choices against published evidence — with a report designed for that capability alone, not a template. They span hiring, compliance, education, operations and personal skill.
Browse the catalogue →Methodology: Forty original exercises across five formats: sixteen single-choice severity calls, each describing one defect on one unit and offering the same four severity options; eight select-every-one-that-applies exercises on sampling results, tolerances and repeated defects; seven true-or-false claims about sampling and measurement; five match-the-following exercises; and four ordering exercises on the inspection, measurement, hazard-handling and close-out procedures. One response instruction is declared for the whole instrument and it is a KNOWLEDGE instruction: what severity the described defect carries and what the described result permits, never what the respondent would feel or prefer. Construct statement: this measures whether a person's severity calls on described garment defects line up in SHAPE with the calls of the people who set the standard, together with what a sampling result permits, how a measurement sits against its tolerance, and what a repeated defect points to; it does not measure eyesight, inspection speed, sewing or pattern skill, knowledge of any buyer's private manual or any country's factory law, how a person behaves under a shipment deadline, or their honesty, and it is not an inspector's certification. Scoring is profile-match correlation (C3). The sixteen severity options are printed in a fixed ladder, so the option a respondent picks IS the severity level they assigned, and their sixteen levels form one vector. The headline is the Pearson correlation between that vector and the expert key vector, corrected against the expected correlation of a respondent whose answer to every severity call is drawn from that call's declared option marginals, by a seeded Monte Carlo of four thousand draws. Zero on the published scale is that ordinary respondent and one hundred is the key exactly; the expected correlation at the declared marginals is printed on the report as the zero of the scale. A correlation is invariant to linear transformation, so elevation and scatter drop out of the headline with no ipsatisation, and the report says so in words. Because the correlation discards level information entirely, two further figures are printed separately and labelled DESCRIPTIVE rather than scored: the signed severity lean, the mean of the respondent's levels minus the mean of the key's, which says whether they call defects generally harsher or generally softer than the key; and the spread ratio, their standard deviation over the key's, which says whether they use more or less of the ladder. The report states in plain words that the correlation cannot see either of them. Exact agreement, the share of the sixteen calls that matched the key level for level, is printed beside the correlation as the level counterpart to the shape and is chance-corrected against the same declared marginals. A respondent who gave the same severity level to every call has no shape, and the correlation of a flat vector would be our arithmetic rather than their answer: the match is REFUSED in words, the refusal occupies the place the figure would have taken, and the flatness test is run on the raw picks. Sixteen calls clears the fifteen-item floor a per-person correlation needs; a sitting with fewer than fifteen severity calls answered carries no correlation, and the page says that below fifteen a per-person correlation is unstable rather than merely imprecise. The report is a confidence ribbon (D3). Every reported figure is drawn as a band and never as a point: an outlined box at plus or minus 1.96 standard errors of measurement with the point estimate as a faint tick inside it, a hatched inner box at plus or minus one standard error, and the plain sentences that about two chances in three of a re-inspection would land inside the inner band and about nineteen in twenty inside the outer one. The bands carry a visible outline and no gradient, because a gradient bands on a laser printer. Where two areas' bands overlap the page prints that they are not distinguishable at this length instead of ranking them. The other twenty-four exercises are corrected against the declared prior on every option, claim, pair and position and feed three further areas beside the severity calls: sampling and acceptance decisions, measurement and tolerance, and disposition and root cause. Reliability is reported as omega and never as alpha, assumed from the answered count and a declared average inter-item correlation of .24, printed unrounded and stated in the open as an assumption that observed data will replace. An area with fewer than eight answered exercises or an omega under .70 carries a three-way classification and no number and no percentile, and the reliability table prints that refusal rather than performing it silently. No true-or-false prior in the file is .50/.50. Refusal rules: an unanswered exercise leaves the numerator, the denominator and the chance term together, never scoring zero; an empty sitting scores exactly zero on every figure and is reported as nothing rather than as a floor; a sitting with fewer than twenty-four of the forty exercises answered is not reported at all, and the refusal is printed where the ribbon and the figures would have been rather than beside them. A careless-responding count is computed for the operator and is never shown to the respondent as a judgement. Legal: no brand, buyer, retailer, mill, machine, software or inspection-scheme name appears anywhere in this instrument, and no competitor is named. Every garment is invented and no real lot was inspected. Every rule is stated inside the exercise that applies it, as the standard your team uses, so no exercise depends on one country's factory law or one buyer's private manual. This is not an inspector's certification, does not qualify anybody to release a shipment, and confers no authority over any lot. Sources drawn on: ISO 2859-1, Sampling procedures for inspection by attributes - Part 1, for the public structure of an attribute sampling plan (the standard and its designation belong to the International Organization for Standardization; no affiliation with or endorsement by that body is implied, and no table, figure or clause of it is reproduced here); Dodge and Romig, Sampling Inspection Tables (1944), for the origin of acceptance sampling; Duncan, Quality Control and Industrial Statistics (5th edition, 1986), and Schilling and Neubauer, Acceptance Sampling in Quality Control (2nd edition, 2009), for attribute-plan theory and the two risks; Montgomery, Introduction to Statistical Quality Control (7th edition, 2012), for the operating characteristic curve and why a passed lot is not a defect-free lot; Shewhart, Economic Control of Quality of Manufactured Product (1931), for the distinction between a signal and ordinary variation; Juran and Godfrey, Juran's Quality Handbook (5th edition, 1999), for the public minor-major-critical classification of defect seriousness; Ishikawa, Guide to Quality Control (1976), for the cause-and-effect reasoning behind the root-cause exercises; Reason, Human Error (1990), for the separation of errors of commission from errors of omission; Mehta, An Introduction to Quality Control for the Apparel Industry (1992), and Kadolph, Quality Assurance for Textiles and Apparel (2nd edition, 2007), for garment inspection practice and the severity conventions; Saville, Physical Testing of Textiles (1999), for measurement and tolerance practice; Cooklin, Garment Technology for Fashion Designers (1997), for measurement points and construction; Cronbach and Gleser, Assessing similarity between profiles (1953), for the decomposition of profile similarity into elevation, scatter and shape, which is why the correlation is the headline and the lean and the spread are printed apart from it; McDonald, Test Theory: A Unified Treatment (1999), for omega; Haladyna, Downing and Rodriguez (2002), for the item-writing and cue-control rules; and Gollwitzer and Sheeran, Implementation intentions and goal achievement (2006), for the single if-then action the report ends on. All exercises are original works written for this instrument. No commercial instrument's items or name are used or implied, and no standard's tables are reproduced.