Stitching and Thread: What QC Inspectors Look For in Bags

What exactly does an inspector look for in the stitching of a handbag? Five things, in a fixed order: stitch density counted against the drawing, whether load-bearing seams carry their specified backtack and reinforcement, whether any stitch has been skipped or broken, the length of exposed thread ends, and whether the seam holds under a measured pull rather than looking tidy. Everything else about stitch quality is downstream of those five. The numbers that matter are unglamorous and specific — 8 to 12 stitches per inch depending on panel, a #18 needle in most leather work, 40s polyester or bonded nylon thread, backtacks of 10-15 mm at every start and stop, thread ends trimmed to 3 mm or less — and the reason each one exists is that a seam which fails in the field almost always failed one of those five checks first. This page is written as a checklist you can hand to an inspector, with the reason attached to every line.

The stitch parameters worth writing onto a handbag drawing

Read the Drawing Before the First Unit

Every check below presumes one thing: the drawing states the parameters. An inspector who arrives at a lot with no stated stitch density can only report that the stitching looks uneven, which is an opinion and will be answered with another opinion. Every number in this checklist is therefore a number you supply, and the checklist itself is only as strong as that page of the drawing. If your specification currently says durable construction and nothing else, that is the actual defect.

The second reason to read the drawing first is that stitch parameters vary by panel on the same bag. A body panel in 1.4 mm leather at ten stitches per inch looks right and behaves well. The same density on a 0.6 mm garment leather creates perforation that tears under load, which is why light leathers run eight. One density for the whole bag is therefore wrong for at least half of the panels on almost every bag.

Check One: Stitch Density, Counted Not Guessed

Stitch density is measured, not estimated. Put a rule on the seam, count the stitches in one inch, and compare against the drawing band rather than against memory. On an 80-unit sample from a 501-1,200 piece lot under ISO 2859-1 General Level II, density should be checked on all units because it takes about four seconds, and any unit falling outside the stated band is a countable instance.

• Count in one inch, twice. Take two adjacent inches; a seam that jumped from ten to fourteen in the space of an inch usually indicates the machine slipping rather than a change of operator.
• Check both ends of the seam. Density drift concentrates at starts and stops where the operator accelerates.
• Compare the visible line with the reverse. A machine set correctly produces the same pitch on both faces; unequal pitch signals a tension problem that will also weaken the seam.
• Record the worst unit, not the average. Averages are how systematic problems pass inspection.

Why density matters mechanically is worth understanding rather than memorising. Each stitch hole is a perforation. Too few stitches and the load concentrates on each hole, tearing through the material; too many and the holes coalesce into a line of weakness that behaves like a perforated tear-off edge. There is a band for every material combination, and both sides of the band produce failures that look completely different in the field.

Check Two: Backtacks and Reinforcement on Load Seams

This is the single most valuable check on the list and the one most often skipped. Every sewn seam has to start and stop somewhere, and at each of those points the thread structure has no continuation holding it. A backtack of 10-15 mm locks the stitch line into the material, and without it the seam unravels progressively from the endpoint under load — often weeks after the bag left the store.

1. Walk every load seam to its endpoints and confirm a visible backtack of the specified length at each one.
2. Check for added reinforcement where specified: a second pass, a box stitch, or webbing inserted into the seam at handle junctions.
3. Confirm the reinforcement is what was drawn. An X-box and a plain single-line restitch are not interchangeable; only one of them distributes the vector loads at a handle root.
4. Look at the thread residue. Where a seam has been unpicked and resewn during rectification, the original holes remain visible under raking light.

That last one catches something specific and valuable: units that were repaired rather than remade. Resewing through existing perforations produces a neat visual result and a measurably weaker seam, because the second pass cannot grip material that has already been punched out. Under raking light at roughly a 15 degree angle, the abandoned holes read clearly. Six named load-bearing joins, and what each one needs, are set out in the nine QC practices page.

Check Three: Skipped, Broken and Doubled Stitches

A skipped stitch is one where the needle passed but the thread did not lock; a broken stitch is one where the thread snapped and was left; a doubled stitch — the reverse problem — appears as two threads lying in one perforation. All three are visible without instruments, and all three concentrate at predictable places: turning corners, crossing a previous seam, and moving from two plies to three.

The diagnosis column is the useful part. A single skipped stitch on one unit is noise. Six skipped stitches concentrated at the same corner across a sample means the machine is deflecting at that thickness transition, and every remaining unit in the lot will carry it. Counting by location rather than by unit converts an inspection result into a production instruction.

Check Four: Thread Ends and Interior Trimming

Trimmed to 3 mm or less, inside and out, is the working standard on most programmes; some brands run 2 mm on visible seams. Exposed thread ends are the most complained-about defect in retail because the customer can see it immediately and cannot un-see it, and it is entirely controllable on the line with nothing more than discipline and sharp scissors.

• Turn lining pockets fully out when checking. Ends hidden in an inverted pocket come out in use and read as a defect then.
• Check under hardware. The area beneath a plate or a foot is the most commonly missed trimming location.
• Look for thread tails left inside the body. A tail of 40 mm inside an unlined body migrates out through the seam over weeks.
• Distinguish trimming from fraying. A trimmed short end is a minor; a frayed end indicates thread of insufficient twist and is a material problem.

Check Five: Seam Strength, Measured Not Inferred

A seam that looks tidy can still be weak, and the only way to know is to pull it. Two test procedures are in general use for sewn seams: ASTM D1683, which measures the force and elongation at failure of a sewn seam in woven fabric, and ISO 13936-2 for resistance to seam slippage under a fixed load. Both are laboratory methods run on coupons rather than finished bags, which is precisely why they belong in pre-approval rather than in-line inspection.

The practical version used during a lot inspection is simpler and still informative: six retained units, a specified load of 25 kg applied statically to each handle junction for one minute, then release and examine. No permanent deformation, no audible crack from the interlining, no stitch-line opening wider than 0.5 mm. That is a pass. What it catches — an under-specified seam allowance, a missing reinforcement, or a thread that is wrong for the ticket — is exactly what field returns are made of.

The Thread and Needle Specification Behind the Checks

Thread choice determines most of what the five checks find. In leather work the common specification is a 40s ticket polyester core-spun thread, sometimes bonded nylon where abrasion resistance is more important than heat resistance. The differences are not cosmetic: polyester holds colour under light better, nylon stretches more and tolerates shock loading, and bonding reduces the fraying that produces visible tail ends three weeks into the product's life.

1. Match needle to stack thickness. A #18 (NM 110) handles most 1.2-1.6 mm leather panels; going too fine produces broken stitches and too coarse produces visible perforations in a line.
2. Change needles on a schedule, not on failure. A dull needle burns thread and raises the local temperature enough to weaken it measurably.
3. Specify the seam allowance. 8 mm is standard and 10 mm on load joins; below 6 mm most stations cannot hold a consistent stitch line.
4. State the thread colour match rule. Either exact match plus one shade darker is acceptable, or match within one commercial shade — pick one and write it.

There is also a compliance dimension that sits behind all of this and is easy to miss. Restricted substance limits apply to thread and to dyed leather alike under REACH Annex XVII, so the fixation of dyestuffs on both components needs to be part of the filing rather than assumed. The Leather Working Group audit protocols include exactly this kind of input traceability, and programmes that already place a tannery requirement on hides usually extend it to thread without difficulty.

Which of the Five Count as Major Defects

Classification should be settled before inspection, because it decides whether the lot fails. Working practice on most handbag programmes runs like this: a missing backtack or a missing specified reinforcement on a load-bearing seam is a critical, since it is a safety-of-function issue and the units cannot be relied upon; skipped or broken stitches on a load seam are major; the same faults on a decorative topstitch are minor; thread ends over 3 mm are minor; and any deviation from stated stitch density is major because it is systemic rather than incidental.

The pattern worth noticing is that location beats appearance. The identical visual fault is a critical on one seam and a minor on another, and the drawing is what establishes which seam carries load. Writing the six named load joins into the specification is therefore what makes classification possible at all, and it takes one line per join. Where defects then get counted and turned into a pass or fail decision is set out in the quality control reference.

Running the Five in Eight Minutes per Unit

Eight minutes is what the five checks take on a medium-complexity bag when done properly, which matters because it determines whether this can run inside a normal sampling plan. On an 80-unit sample that is a bit over ten hours, comfortably one man-day, and inside the USD 220-320 normally quoted for a pre-shipment visit. On a 125-unit sample it is closer to seventeen hours, which is why most programmes either split it across two days or agree a reduced set for a second lot of a proven style.

1. Minute 0-1: read the drawing band for each panel and confirm the assigned inspection acceptance criteria.
2. Minute 1-3: count density on both ends of every seam, worst-unit recording.
3. Minute 3-5: walk load seam endpoints for backtacks and specified reinforcement.
4. Minute 5-7: visual sweep for skipped, broken and doubled stitches, noting location each time.
5. Minute 7-8: thread ends inside and out, pockets turned out, under hardware.
6. Separately, on six units: the 25 kg static pull and the post-release examination.

Doing it in that order is not ceremony. Density first establishes whether the machine was set correctly, and the answer changes how charitably to read everything after it. If density is at the bottom of the band and there are skipped stitches at the same corner, you are looking at one machine in need of adjustment rather than several unrelated problems, and your report should say so in one sentence at the top.

Once you know what to look at, the remaining question is whether to check it yourself or bring someone in, which is dealt with in when an independent inspection visit earns its fee. Either way, the checklist is the same five lines.

The Machine Setting Behind the Count

Stitch density is a symptom; the machine setting is the cause. An inspector who counts ten stitches to the inch has recorded an outcome produced by five settings, and when the count drifts, the setting is where the correction is made. Recording the settings at sample approval is what makes a bulk drift explainable rather than arguable.

Two of those six explain most of what the five checks find. Tension imbalance produces uneven top and bobbin thread and measurably lower seam strength, and a needle kept past its interval raises the local temperature enough to weaken the thread it is sewing with, which is why a tidy seam still fails a 25 kg pull.

How the Same Check Changes by Material

A single stitch density figure on a drawing is not a specification, because the same 8-10 stitches per inch means different things on a hide, on a coated fabric and on a 210D lining. What changes with the material is the needle, the allowance and the fault that appears first, and an inspector working from one number will under-check two of the three.

The practical rule is to write the density per panel on the drawing rather than once for the bag, and to name the needle on the same line. A specification that says NM 110 and 8-10 spi on the body and NM 90 and 9-11 spi on the lining gives an inspector something to check against, and gives the sample room something to set the machine to.

What the Stitching Page of the Report Has to Show

A report that says the stitching is acceptable is worth nothing to anybody who was not in the room. Six elements make the stitching page reproducible, and all six are checkable in a minute by a buyer reading the report a week later.

1. Sample size and code letter. 50 units on 281-500 and 80 units on 501-1,200 under ISO 2859-1 at General Inspection Level II.
2. Counts by panel, not for the bag. A single average hides a systematic fault concentrated on one seam.
3. The worst unit recorded, not the mean. Averages are how systematic problems pass inspection.
4. Photographs with a 1 cm scale. One per complaint category, plus a full-view shot of the unit concerned.
5. Classification applied. Critical, major and minor agreed before the visit rather than argued during it.
6. Accept and reject numbers applied. Three and four on a 50-unit draw, five and six on an 80-unit draw at AQL 2.5.

Keep the file for twelve months. The comparison that matters is not against the standard, it is against the last lot: a programme whose defect count rises from two to five across three runs has a maintenance or turnover problem, and that trend is only visible in a report that recorded numbers rather than verdicts.

Frequently Asked Questions

People Also Ask