Setting Tolerances on Knitted Bags: How a Gap of Three Millimetres Became a Rejected Container, and the Numbers That Prevent It

The Rejection That Started With Three Millimetres

A buyer's tech pack specified a tote at 380 mm long and did not specify a tolerance. The production run measured between 376 and 379 mm at goods-in, comfortably close to the drawing by any reasonable reading, and the shipment sailed. Four weeks later the retailer's own quality check measured the same bags at 371 mm and rejected the delivery.

Nothing had been made wrong. A knitted panel relaxes after knitting, relaxes again after packing under compression, and relaxes once more when it hangs on a fixture. Three stages of movement, none of them a defect, and a specification that named a target without naming the conditions under which the target would be checked.

The buyer's second order carried three additional lines: a tolerance of plus or minus 3 mm, a conditioning period of 4 hours under ISO 139, and a resting period of 24 hours after unpacking before measurement. The same bag, the same machine, and no rejection. The illustration is that the argument was never about three millimetres; it was about when a millimetre counts.

Why a Knit Needs a Different Tolerance Logic

A tolerance on a woven bag is mostly a cutting tolerance: the panel was cut to a size and it stays that size. A tolerance on a knitted bag is a process tolerance, because the panel was formed loop by loop under tension and it continues to change shape after the tension is released. Treating the two the same is the origin of most dimensional disputes.

Three variables drive the movement. Take-down tension on the machine sets how much the panel is stretched while it is being formed. Yarn type sets how much of that stretch is recoverable, with elasticised or textured yarns recovering more than flat ones. And finishing, particularly any washing or steaming, sets how much of the residual strain is released before the bag is packed.

The practical reading for a buyer is that a knitted bag tolerance is a range around a process, not a limit on a part. Writing plus or minus 3 mm is an acknowledgement that the process produces a distribution; writing a single target figure implies a precision that a knit does not have and invites an argument the first time a bag measures 377 mm.

The Tolerance Set: Nine Numbers

Nine numbers is a complete tolerance set for a knitted bag and most tech packs carry two. The first five are dimensional, the sixth is a performance limit, and the last three are the conditions that make the first six testable. Omitting the last three is the single most common cause of a dispute that no tolerance figure can resolve.

The resting period is the row most often omitted and the one that most often explains a rejection. A bag packed flat under compression for three weeks and measured the moment it is unpacked is at its smallest. The same bag measured a day later, having been allowed to lie unrestrained, is measurably larger, and the difference sits within the tolerance band rather than outside it.

When the Bag Is Measured Matters More Than the Number

A measurement is a number taken at a moment, and on a knitted bag the moment decides the number. Three moments are available and they give three different answers: on the machine, at goods-in, and at destination after unpacking and resting. A tolerance written without naming the moment is a tolerance that will be tested at whichever moment suits the person holding the tape measure.

The defensible convention is to measure at goods-in after conditioning and resting, because that is the first point at which both parties can reasonably be asked to agree on a number and the last point at which a remedy is still inexpensive. Measuring on the machine is meaningless to a buyer, and measuring at destination is a conversation about who is right rather than about the bag.

Writing the moment into the specification also settles the shrinkage question. A shrinkage limit of 3 percent means nothing until it says three percent after what: after one domestic wash at a stated temperature, or after the first month on a fixture. Both are legitimate and they are not the same test, so the condition belongs beside the number rather than in a footnote.

Where Tolerance Meets Function

Some dimensions on a bag are not appearance dimensions, and they need a tighter tolerance for a functional reason rather than an aesthetic one. Identifying which is which is the difference between a tolerance set that protects the product and one that merely documents it.

The discipline is to write a one-sided limit rather than a symmetric band wherever the dimension is functional. A pocket that must accept a device has a minimum opening, not a plus or minus figure, and stating it that way removes the possibility of a compliant bag that does not do its job.

Dimensional Tolerance and AQL: Two Different Instruments

Buyers often try to express a dimensional requirement inside the AQL and it does not fit. The AQL is an attribute count: a piece either has the defect or it does not, and the accept number counts pieces. A dimensional result is a variable: every piece has a measurement, and the question is where the distribution sits relative to a limit.

The workable arrangement is to run both and keep them separate. Count attributes at AQL 2.5 with zero critical, 2.5 major and 4.0 minor, and separately measure five pieces per colourway against the dimensional tolerance, with the lot failing if any measurement exceeds the limit or if the mean drifts more than half the band from the target.

The mean drift provision is worth having and is usually left out. Five bags all inside plus or minus 3 mm but all measuring 377 mm against a 380 mm target are telling the buyer that the process has shifted, and the next lot will start producing failures. Half the band is a reasonable trigger for a process check rather than a rejection.

Writing Tolerances Into the Tech Pack

Tolerances belong on the tech pack next to the dimension they govern, not in a general notes section. A note saying all dimensions plus or minus 3 mm is worse than no note, because it applies a body tolerance to a handle placement and a pocket opening, and those three need three different figures.

Body length

380 mm, plus or minus 3 mm

Body width

340 mm, plus or minus 2 mm

Handle width

25 mm, plus or minus 2 mm

Handle placement

Centre line, plus or minus 4 mm symmetric between the two handles

Weight per bag

Plus or minus 5 percent against the approved reference bag

Shrinkage

Not above 3 percent after one domestic wash at 40 degrees Celsius

Conditioning

4 hours at 20 degrees Celsius and 65 percent relative humidity, ISO 139

Resting before measurement

24 hours after unpacking, laid flat and unrestrained

Specimens

5 per colourway from 5 different cartons, mean and spread both recorded

Process drift trigger

Process check when the mean moves more than half the tolerance band from target

One more line completes it and it costs nothing: the remedy. Whether a lot that measures outside tolerance is reworked, accepted with an allowance, or rejected, and who decides. Without it, a dimensional failure becomes the same open-ended conversation as any other rejection, at the point in the programme where the buyer has least time.

Reviewing Tolerances After the First Season

A tolerance set is a first guess written before any production data exists, and it should be reviewed once there is data. After a season, a buyer knows the actual spread the process produces, and the tolerance can be tightened where it was too loose or widened where it was generating rework with no benefit to the product.

The review is straightforward: collect the measurement means and spreads from every lot, plot them against the tolerance band, and look at where the distribution actually sits. A band that the process fills entirely is too tight, because it will start producing failures from normal variation. A band the process uses a third of can be tightened without cost.

Frequently Asked Questions

People Also Ask