What is a defect catalogue?

A controlled list of every defect type on a part: its name, a definition, a class, where it occurs, how to detect it, and the accept and reject limits, with boundary samples for anything that needs judgement. It makes every inspector judge the same defect the same way.

How do you decide whether a defect is critical, major or minor?

Ask in order: could it hurt someone using, maintaining or relying on the product (critical); if not, is it likely to cause failure or materially reduce use (major); otherwise it is minor. These follow the classification of defects in MIL-STD-105D, section 2.1, the public military standard that ANSI/ASQ Z1.4 grew out of.

What is a boundary sample?

A real part, or a photo or drawing when no part exists, that shows the limit of a defect: the worst condition still accepted. It sits next to a reject just beyond the limit, carries an ID and an approval date, and is re-approved on a schedule so it does not drift. Automotive suppliers with appearance items should also read IATF 16949 clause 8.6.3 and their customer's requirements.

How do I check that my inspectors agree?

Run an attribute agreement check: 20 to 30 parts with a reference decision, about half of them rejects and several near the limit, each judged twice by each inspector in a random order. Read effectiveness and the miss rate first; the AIAG guide, as reported by SPC for Excel, calls 90% effectiveness, 2% misses and 5% false alarms acceptable. A full study uses more parts, such as the 50 parts, 3 inspectors and 3 trials in the AIAG manual's example.

How is a defect catalogue different from a visual standards board or a control plan?

A visual standards board shows a few good and bad photos at one station; the catalogue is the controlled master list it should be made from. A control plan says what is checked, how often and what to do when a check fails, and can point to a catalogue row, but it does not define the defect or its limit. The defect check sheet counts defects using the catalogue's codes.

Free template

Updated October 9, 2026

Defect catalogue template with boundary samples

A defect catalogue (also called a defect library or appearance standard) lists every defect type on a part with a name, a definition, a class (critical, major or minor) and the limit where accept ends and reject starts, so every inspector judges a defect the same way. Where a limit cannot be measured, a boundary sample shows it: a real part at the limit, kept next to a reject just beyond it. This template has the catalogue, a boundary sample card with accept, borderline and reject panels and a zone diagram, a defect log that takes its codes from the catalogue, a summary with defects per unit and a Pareto of defect codes, and an attribute agreement check that shows whether your inspectors agree with each other and with a reference decision. The class questions are paraphrased from MIL-STD-105D, the public military sampling standard that ANSI/ASQ Z1.4 grew out of. To count defects without defining them, use the defect check sheet; to rank where defects are made and caught, use the QA matrix.

Free to use: print it, copy it and edit it for your team. Enter your name and work email once to download.

Page 1 of the defect catalogue template — Defect definitionsCritical, major, minor with the reasonZones A, B and CAccept and reject limitsBoundary sample cardDefect log linked to the catalogueDefects per unit and ParetoInspector agreement checkWorked example

When to use it

When to use a defect catalogue

When a new part goes into production and inspectors need to know what counts as a scratch, a dent or a crack before the first shift.
After a customer complaint or a returned lot, when your inspection accepted what the customer rejected.
When two shifts reject very different amounts of the same product, or the scrap bin fills with parts that look fine.
When the check sheet, the NCRs and the QA matrix each use their own names for the same defect.
Before you train a new inspector or run an agreement check, so there is one reference to train and check against.

How to fill it in

1
Start from the defects you actually have
Pull the last few months of check sheets, NCRs, scrap tags and complaints, count each defect type and write the top 10 first. Add the rest later. A short catalogue people use beats a long one nobody reads.
2
Name and define each defect
One name and one code per defect, used on the floor, in the log and in reports. The definition says what the defect is, for example a line mark in the paint, not how bad it is.
3
Set the zones
Split the part by how much each surface is seen: zone A for the face the customer looks at, B for sides seen at an angle, C for surfaces hidden in use. The same mark can be fine in zone C and a reject in zone A.
4
Pick the class with three questions
Could it hurt someone using, maintaining or relying on the product? Critical. If not, is it likely to make the unit fail or materially reduce its use? Major. Otherwise, Minor. Write down the answer that decided it: the Excel flags a class that does not match its reason.
5
Write limits two inspectors would apply the same way
Put a number on every limit you can: length, depth, diameter, count per face, distance apart, and the zone. Say how to look: distance, light, angle, seconds per face and which gauge. No noticeable scratches fails this test; zone A, 1 scratch up to 5 mm with no plastic showing, seen at 450 mm, passes.
6
Make a boundary sample for anything you cannot measure
A real part at the limit (the worst condition still accepted) with a reject just beyond it, labelled with an ID, approved by quality and by the customer where they require it, kept at the station, and re-approved on a date you set. The card holds a photo or sketch with a scale.
7
Check that inspectors agree
Pick 20 to 30 parts, about as many rejects as good ones, with several close to the limit. A quality engineer sets the reference decision. Each inspector judges every part, accept or reject, in a random order, twice, without seeing the reference or anyone else's answers. Enter the results in the Excel and read them with the table below.
8
Review every quarter
Review the catalogue every three months (an example interval the Excel uses) and after any complaint, new defect or design change. Retire codes nobody logs, split codes that hide two causes, and run the agreement check again when a limit changes or a new inspector starts.

Reading an agreement check

What each figure in the inspector agreement check means

The measures follow Minitab's attribute agreement analysis. The guide values are the AIAG MSA manual's (3rd edition) as reported by W. McNeese of SPC for Excel in 2010; the manual itself was not read. A 20-part check is a screening check: one part moves each percentage by 5 points. The AIAG manual's own example uses 50 parts, 3 inspectors and 3 trials.

Figure	What it counts	How to read it
Agrees with self	Parts where all of one inspector's trials gave the same answer	Low means the inspector is guessing near the limit. Check the limit wording and the boundary sample first.
Effectiveness	Parts where every trial matched the reference decision	Read this first. Guide: 90% or more acceptable, 80% to 90% marginal, under 80% unacceptable.
Miss rate	Ratings of reject parts that said accept, out of all ratings of reject parts	These reach the customer. Guide: 2% or less acceptable, over 5% unacceptable.
False alarm rate	Ratings of good parts that said reject, out of all ratings of good parts	These cost scrap and rework. Guide: 5% or less acceptable, over 10% unacceptable.
Between inspectors	Parts where every rating from every inspector agreed	Agreement is not correctness: three inspectors can all accept a reject.
All vs reference	Parts where every rating matched the reference	The share of parts the whole team gets right every time.

Agrees with self
What it counts
Parts where all of one inspector's trials gave the same answer
How to read it
Low means the inspector is guessing near the limit. Check the limit wording and the boundary sample first.
Effectiveness
What it counts
Parts where every trial matched the reference decision
How to read it
Read this first. Guide: 90% or more acceptable, 80% to 90% marginal, under 80% unacceptable.
Miss rate
What it counts
Ratings of reject parts that said accept, out of all ratings of reject parts
How to read it
These reach the customer. Guide: 2% or less acceptable, over 5% unacceptable.
False alarm rate
What it counts
Ratings of good parts that said reject, out of all ratings of good parts
How to read it
These cost scrap and rework. Guide: 5% or less acceptable, over 10% unacceptable.
Between inspectors
What it counts
Parts where every rating from every inspector agreed
How to read it
Agreement is not correctness: three inspectors can all accept a reject.
All vs reference
What it counts
Parts where every rating matched the reference
How to read it
The share of parts the whole team gets right every time.

A filled-in example

Illustrative, not a benchmark

An example (illustrative, made up): a wall-mounted control unit with a molded housing, a painted front cover, a display window, a rating label, 4 cover screws and a cable with a latching connector.

The example's 92 defects, and how well the inspectors agree

Rejects passed as good (miss rate), 20 reject ratings each

Inspector A, 95% effective2 of 20 (10%)misses rejects
Inspector B, 95% effective0 of 20 (0%)no misses
Inspector C, 75% effective7 of 20 (35%)misses rejects

The two scratch defects are 53 of the 92 (57.6%). Inspector A was 95% effective and still passed a reject twice, so the team read the miss rate first and added a reject sample with plastic showing to the paint scratch card.

The catalogue holds 10 defects: 1 critical (damaged cable insulation, because a cut jacket can expose a live conductor), 4 major (missing or loose screw, cracked housing, connector not latched, unreadable rating label) and 5 minor (window scratch, paint scratch, sink mark, particle under the window, crooked label).
The paint scratch limit changes by zone: zone A accepts 1 scratch up to 5 mm with no plastic showing, zone B up to 3 scratches up to 10 mm, zone C anything unless plastic shows over more than 20 mm.
Ten days of final inspection: 2,440 units and 92 defects beyond their reject limits, on 86 units. That is 92 / 2,440 = 0.0377 defects per unit (3.77 per hundred units), and 86 / 2,440 = 3.52% of units with a defect: 0 critical, 13 major, 79 minor.
The Pareto: paint scratch 31 (33.7%), window scratch 22 (57.6% cumulative), particle under the window 12 (70.7%), sink mark 9 (80.4%).
The agreement check, 20 parts, 3 inspectors, 2 trials: inspector A was 95% effective but passed one reject twice (2 of 20 reject ratings, 10%); B was 95% effective with 1 false alarm in 20 (5%); C was 75% effective with 7 of 20 misses (35%). All three agreed on 14 of 20 parts (70%).

The two scratch defects are 53 of the 92 (57.6%). Before changing the process, the team added a reject sample with plastic showing to the paint scratch card, because A and C had passed a 5 mm scratch with plastic showing, and coached inspector C with the samples.

Common mistakes

Limits written as adjectives
No noticeable scratches or minor dust acceptable gets a different answer from each inspector and each shift. A limit with no number and no boundary sample is not a limit; the Excel flags it.
Classing by how annoying the defect is
A cosmetic defect a customer complains about is still minor unless it affects use or safety. Keep the class for how serious the defect is, and tighten the zone A limit if the customer cares about the front face.
Boundary samples nobody controls
A sample handled for a year, faded or lost proves nothing. Label it, keep it in one place, record each check, and re-approve it on a date.
One limit for the whole part
The same scratch can be a reject on the front and invisible on the back. Without zones, inspectors either reject good parts or pass bad ones.
Checking only that inspectors agree with each other
Three inspectors can agree and all be wrong. Compare every inspector with a reference decision set by quality, and look at misses first: they are the ones that reach the customer.

Download the template

Free to use: print it, copy it and edit it for your team. Enter your name and work email once to download.

Run this template in LeanSuite

In LeanSuite defects are logged as tags and ranked in the QA Matrix, and digital checklists with number limits, required photos and signatures build the checks into every task.

See it in a 45-minute demo

In the lean glossary

Free tools

Guides

FAQ