Defect Dictionary for Bags: Photographic Guide to Major and Minor Defects

The most useful quality document in most programmes is not the sampling plan — it is one nobody writes. The plan tells you how many units to draw; the dictionary tells you what counts once you have drawn them. Get the second one wrong and the first stops mattering, because two inspectors looking at the same fifty units will produce genuinely different reports.

This is not theoretical. Pull any three reports for the same style from three different visits and compare how the same condition is described. "Thread ends", "loose threads" and "untrimmed threads" may or may not be the same finding; "colour deviation" might be anything from invisible to unacceptable. The reports are not comparable, which means nothing is being tracked even though everything is being recorded.

What Is a Defect Dictionary Actually For?

Three jobs, in order of importance. It makes findings comparable across visits, so a rising count means something. It makes the sampling plan meaningful, because an acceptance number is meaningless without agreement about what is being counted. And it makes conversations short, because most disputes are about definitions rather than about the goods.

It also has a quiet effect on production. When the line supervisor can see the thirty faults that will be counted, with photographs, those thirty get addressed before anything else does. A dictionary is a specification that everyone on the floor can actually read.

And it changes the economics of delay. Without a dictionary, a failed lot sets off a conversation about definitions before anybody gets to the corrections; with one, the conversation starts at "sort or remake". Against a 35-50 day bulk window, the days spent agreeing what a fault is called are days that could have gone into fixing it.

How Do You Write an Entry an Inspector Can Actually Use?

Write it as instruction rather than as description. "Stitch density below eight per inch measured over three inches at any load-bearing seam" can be executed by anybody on any floor. "Poor stitching" cannot be executed by anybody, and inspectors will each quietly invent their own version of it.

Avoid adjectives wherever a measurement exists. "Lightly soiled", "slightly out of line" and "noticeably darker" all shift with who is reading them. Replace each with a limit — a stated length, distance or grade — and the entry becomes enforceable. Where a measurement genuinely does not exist, keep the adjective but attach a reference image and say so.

Finally, write each entry in the language the inspector works in, or provide both versions side by side. An entry translated badly at the point of use is the same as an unclear entry, and it is a common failure in programmes sourced across languages.

How Should Each Entry Be Structured?

Five fields, consistently filled: the name, where it occurs, how it is detected, what counts as one occurrence, and the class. The fourth is the one people forget and it resolves a surprising amount — is a seam fault counted once per unit or once per occurrence along the seam? Agreeing this removes an argument the inspector currently resolves alone.

How Do You Decide Which Class an Entry Belongs To?

Ask what the customer does when they notice it. If they return the bag, it is a major. If they notice and keep it, it is a minor. If it creates a hazard or breaks a rule, it is critical, whatever anybody notices. That is the whole test, and applying it consistently to every entry is what makes a dictionary coherent rather than arbitrary.

The interesting cases sit where "the customer" is ambiguous. A fault visible at the point of sale behaves differently from one visible at home; a fault on the outer face differs from one inside a lining. Write down which customer and which moment the dictionary assumes, then apply it consistently. Most inconsistency in real dictionaries comes from answering that question differently for different entries.

Which Defects Belong in the Critical Class?

A short list, deliberately. Everything that could make the product unsafe or unlawful, plus anything the destination market specifically controls. Critical findings carry zero acceptance: one occurrence is enough to fail the lot regardless of how good the rest of the sampling looks.

What Does the Major List Contain?

Everything that loses the sale or loses the function. It is the longest list in most dictionaries and it is the one that actually moves the accept and reject numbers, because majors are what AQL 2.5 is set against.

Two refinements make the list behave better in practice. First, decide how several majors on one unit are counted — per unit or per occurrence. Most programmes count the unit once in the major column and note the occurrences, because counting every occurrence would make a single bad unit look like a lot-wide problem. Second, decide what happens when one fault could be recorded under two entries; give one of them precedence and say which.

What Belongs on the Minor List?

Departures from specification that a customer could see and would keep the bag anyway. Minors are counted at AQL 4.0 under this programme, which means they accumulate more tolerance than majors by design — but only if the list is honestly maintained.

How Do You Keep the Dictionary From Growing?

It grows for the wrong reason: somebody sees a problem, adds an entry, and the list accumulates exceptions nobody revisits. Three hundred entries later, nobody applies them consistently and the document has quietly stopped being useful.

Two habits prevent this. Review the list at the end of each programme rather than continuously during it, adding only what actually occurred. And retire entries — if a named fault has never been recorded in four lots, ask whether it is being looked for or whether it simply does not arise. Retiring is healthy; nobody should treat the dictionary as permanent.

How Should the Photographs Be Managed?

One image per named fault, taken as deliberately as any other reference: consistent background, consistent scale, a stated light source, and the location marked in the frame where position matters. These are reference images rather than report images, so clarity beats convenience — a phone picture under the right conditions serves perfectly well.

Version everything. Put a number, a date and the order reference on each page, and send the current version with every booking. Half of the value of a dictionary comes from both sides being able to prove which version applied when the goods were inspected.

What Happens When a Fault Is Not in the Dictionary?

It happens constantly and needs a route. The sensible one is to record the finding as observed, without a class, photograph it, and refer it for classification rather than letting the inspector assign one. Add it to the next revision with the class everybody agreed.

What does not work is silence — the fault not being written down because it is not in the list. A dictionary exists to describe the product, not to limit what can be noticed. Say this explicitly when you hand it over, because inspectors are trained to work to lists.

Set a turnaround for that referral. A finding awaiting classification should not sit unresolved while the packing continues around it, and naming who answers and within how long keeps the exception from becoming a delay.

Store the record with the lot information rather than in a folder of photographs. A finding attaches to a lot, a line and a date, and that is how it will need to be found two years later.

How Does the Dictionary Reach the Inspection Company?

As an attachment to the booking, in the same file every time, referenced by version. Do not paste extracts into an email body; it guarantees that two people are working from different versions within a fortnight.

Ask for confirmation that the checker working your programme has read it, and ask what happens when the named inspector is unavailable. Substitute inspectors working from a list they have not read are one of the quietest sources of inconsistency there is.

Where Does This Fit Alongside the Sampling Plan?

Beside it, not after it. The plan sets how many units are drawn and what counts end the discussion; the dictionary sets what those counted things are. A programme can quote AQL 2.5 with complete precision and still not have defined a single defect — which is roughly where a great many programmes sit.

Read together they close the loop. The plan says reject at four majors in fifty units; the dictionary says what a major is. Neither document means much on its own, and the pair of them is what makes a report enforceable rather than merely informative.

Frequently Asked Questions

People Also Ask