How to Reduce Rejection Rate & DHU in Your Garment Factory — From 8% to Under 3%

Our rejection rate was 8.2% in the first month. The buyer said anything above 3% means they find another factory. We had 30 days to fix it.

This was not a new factory learning curve problem. We had operators with 5-10 years of experience. The machines were serviced. The fabric was acceptable quality. Yet one in every twelve garments coming off our line had a defect serious enough to fail inspection. Some garments had two or three defects. Our DHU — defects per hundred units — was actually 12.4, meaning every hundred garments had 12.4 individual defects that needed to be found and fixed.

The problem was not the operators. It was not the machines. It was that we had no system for catching defects where they happen, when they happen. We were doing what most garment factories do: sew everything, inspect at the end of the line, find the problems after they have already consumed 100% of the labor, and then send garments back for rework. It is the most expensive possible way to handle quality.

Here is how we fixed it.

What is DHU, and Why It Matters More Than Rejection Rate

Most factory owners track rejection rate: the percentage of garments that fail final inspection. But rejection rate is a blunt instrument. A garment can fail for one defect or five defects, and the rejection rate counts it the same way. A factory with a 5% rejection rate might actually have a very different quality problem than another factory with the same 5% rate.

DHU — Defects per Hundred Units — is the sharper metric. It counts every defect, not every garment.

Example: You inspect 500 garments. You find 47 defects across those garments (some garments have more than one defect). DHU = (47 / 500) × 100 = 9.4 DHU. The rejection rate might be 6% if 30 garments failed inspection, but the DHU tells you the true defect density is 9.4 per hundred units.

Why does this matter? Because DHU tells you how many problems you need to fix, not just how many garments you need to send back. A factory with 5% rejection and 5.2 DHU has simple, single-defect problems. A factory with 5% rejection and 14 DHU has garments with multiple defects each — a completely different problem requiring a different solution.

Industry Benchmarks

The Top 10 Defect Types — From Real Data

When we started tracking defects systematically, we wanted to know exactly where our 12.4 DHU was coming from. Not "quality is bad" — that is useless. We needed to know which specific defects, on which operations, by which operators.

After two weeks of inline and end-line inspection data, here is what the breakdown looked like:

Look at the top three: skip stitch, uneven stitch, and puckering. Together, they account for 51% of all defects. All three are primarily machine-related problems with an operator skill component. This was our first major insight: more than half of our quality problems were solvable by fixing machines, not retraining people.

Root Cause Analysis — It Is Usually Not the Operator

When quality is bad, the instinct is to blame the operator. They are the last person who touched the garment before the defect was found. But after tracking defect data for a month, we found that the root causes break down into three categories, and only one of them is operator skill.

1. Machine Problems (40-50% of defects)

The machine is the most underestimated source of defects. A needle that is 0.2mm too thick for the fabric creates puckering on every piece. A feed dog with uneven pressure causes SPI (stitches per inch) variation on every seam. A hook timing that is off by a fraction creates skip stitches that seem random but are actually consistent.

The problem is that machine issues create defects that look like operator errors. A supervisor sees skip stitches and tells the operator to be more careful. The operator tries harder but the skip stitches continue because the machine timing is off. The supervisor gets frustrated. The operator gets demotivated. Quality does not improve. Nobody checks the machine.

2. Material Problems (20-30% of defects)

Defects that originate at cutting or even at fabric receipt but only become visible at sewing or finishing. Shade variation happens because rolls from different dye lots were cut together. Measurement problems happen because the fabric was not relaxed before cutting, so pieces shrink after sewing. Stains happen because fabric was stored on a dirty floor.

The critical point: these defects are already in the garment before the sewing operator touches it. No amount of operator training fixes a cutting problem. You have to catch material defects upstream, at fabric inspection and cutting audit, or you will pay for them at sewing and finishing.

3. Operator Skill (20-30% of defects)

Yes, operator skill matters. But it matters less than most factory owners think, and the skill problems are usually very specific. It is not "bad operator." It is "this specific operator struggles with collar curves on single needle because she was trained on overlock and never practiced curved stitching." The fix is targeted training on the specific defect, not a general quality lecture.

7 Steps to Reduce DHU Below 3%

These are the seven changes we made, in the order we made them. The order matters. Each step builds on the previous one.

Step 1: Move to Inline Inspection

This was the single biggest change. We stopped waiting for end-line inspection to find defects and put a roaming quality checker on the line. Her job: check 7 pieces out of every bundle at 3 critical operations (the ones producing the most defects). Not at the end of the line. At the operation itself.

The impact was immediate. When an operator produces a skip stitch and hears about it within 2 minutes, she fixes the problem on the next piece. When she hears about it 3 hours later at end-line inspection, she has already produced 150 more pieces with the same defect. Inline inspection turns a 1-piece problem into a 1-piece problem. End-line inspection turns a 1-piece problem into a 150-piece rework nightmare.

Our DHU dropped from 12.4 to 7.8 in the first week of inline inspection alone. Not because the operators suddenly got better. Because we started catching and correcting problems 10 minutes after they started instead of 3 hours later.

Step 2: Fix the Machines First

We pulled the mechanic off his normal repair schedule for one full day and had him go through every machine on the line that was producing the most defects. The checklist was simple:

• Thread tension: upper and lower within spec for the fabric weight
• Needle: correct type and size for the fabric (ballpoint for knits, sharp for wovens)
• Feed dog: even pressure, no teeth damage, correct height
• Hook timing: synced with needle at correct position
• Presser foot: correct type for the operation, pressure set correctly
• Oil: clean, no leaks, mechanism lubricated

Out of 40 machines, 14 had at least one setting that was out of spec. Three machines had damaged feed dogs that nobody had noticed. Two had hook timing issues that had been producing intermittent skip stitches for weeks. After fixing all 14 machines, our skip stitch rate dropped by 60% and puckering dropped by 45%. One day of mechanic work eliminated nearly a third of our total defects.

Step 3: Track Defects by Operator and Operation

General defect counts are useless for solving problems. You need to know: which operator, which operation, which defect type, and when. Once we started tracking at this level, patterns jumped out.

Operator 7 was producing 80% of the puckering defects on side seams. It turned out her machine's presser foot pressure was too high for the lightweight fabric we were running, and she was compensating by pulling the fabric, which made it worse. We adjusted the pressure and gave her a 10-minute refresher on fabric handling for lightweight material. Her puckering rate went from 8% to under 1%.

Operator 22 had excellent quality on overlock operations but terrible quality on single needle. She had been moved to single needle to fill a vacancy. Her skill was overlock. We moved her back and put a single-needle-trained operator in that station. Quality on both stations improved instantly.

Without operator-level data, both of these fixes would have been invisible. You would just see "high puckering" and "high broken stitch" on the daily report and have no idea where to start.

Step 4: Train on the Specific Defect, Not General Quality

We stopped doing general quality training sessions. They are a waste of time. Standing in front of 40 operators saying "quality is important" changes nothing. Everyone already knows quality is important. They do not know why their collar attachment is producing raw edges.

Instead, when the defect data shows that a specific operator has a recurring problem with a specific defect, the line supervisor spends 5 minutes at that operator's machine demonstrating the correct technique for that specific operation. Not in a training room. Not at the end of the day. Right now, at the machine, on the actual fabric.

Five minutes of targeted, in-context training is worth more than an hour of classroom instruction. The operator can immediately apply what she learned on the next piece. She can feel the difference in the fabric handling. The supervisor can watch and correct in real time. This is how skills actually transfer on a factory floor.

Step 5: Inspect Fabric Before Cutting

This is the step that most sewing factories skip because they think it is not their problem. If you are a CMT factory, the buyer or fabric supplier sends you the fabric. You cut it and sew it. If the fabric has defects, that is the supplier's problem, right?

Wrong. It is your problem because your operators are sewing defective pieces, your quality checkers are rejecting them, and your rework team is spending time on garments that were doomed from the start. A hole in the fabric that makes it through cutting and into a sewn garment has consumed cutting time, sewing time, thread, and inspection time before anyone catches it.

We implemented fabric inspection using the 4-point system on every incoming roll. It adds about 15 minutes per roll, which is nothing compared to the rework time it saves. In the first month, we rejected 3 rolls out of 40 that would have entered production with defect densities above the acceptable threshold. Those 3 rolls would have produced approximately 200 defective garments. At an average rework cost of $0.40 per garment, that is $80 in rework avoided for 45 minutes of inspection time.

Step 6: Measurement Audit at Cutting

Measurement defects are the most expensive defects because they often cannot be fixed by rework. If a garment is 2 cm short in body length because the fabric shrunk after cutting, you cannot add fabric back. The garment is either downgraded or rejected entirely.

We added a measurement check on the first 5 pieces of every new lot immediately after cutting, before bundles go to the sewing floor. The cutter measures body length, chest width, sleeve length, and collar width against the spec sheet. If any measurement is outside tolerance, production stops until the issue is resolved — which is usually a marker adjustment or fabric relaxation step.

This 10-minute check at cutting prevents measurement defects that would otherwise show up at final inspection on 500+ garments. We caught two lots with measurement issues in the first month. Both were fixable at cutting by adjusting the marker. Neither would have been fixable after sewing.

Step 7: Visual Standards at Each Workstation

The simplest change that had a surprisingly large impact. We attached a sample piece to every sewing machine showing exactly what the finished operation should look like. Not a photograph. An actual sewn piece, attached with a clip to the side of the machine at eye level.

When an operator is unsure about stitch placement, seam allowance width, or topstitch distance, she looks at the sample. No need to call the supervisor. No need to guess. The standard is right there, 30 centimeters from her hand. This is especially valuable in the first few hours of a new style when everyone is still calibrating.

We also added a small card next to each machine showing the three most common defects for that operation and how to prevent them. For the overlock station: "Check for skip stitch every 10 pieces. If found, call mechanic immediately — do not continue sewing." For the single needle station: "Maintain 1cm seam allowance. Use the edge guide, not your eye."

The Economics of Quality

Let me put numbers on why quality improvement has the highest ROI of any factory investment.

The cost of rework is 3-5x the cost of the original operation. Here is why. The original side seam operation takes 0.4 minutes and costs $0.02 in labor. Reworking that same seam requires: identifying the defect (0.5 min), transporting the garment back to the operator (1 min), unpicking the defective stitching (2-3 min), re-sewing (0.4 min), and re-inspecting (0.5 min). Total rework time: 4-5 minutes. That is 10x the original operation time. Even at a conservative 3x multiplier (some steps overlap or are faster), rework is catastrophically expensive relative to doing it right the first time.

Now scale that up. Say your factory produces 2,000 garments per day with a 5% rejection rate. That is 100 garments per day going to rework. If each garment has an average of 1.5 defects and each defect costs $0.40 in rework time, that is $60 per day in rework labor alone. That is $1,560 per month or $18,720 per year — purely wasted labor that adds zero value.

But it gets worse. A garment rejected at final inspection has already consumed 100% of its production labor. If your total CMT cost is $1.50 per garment and 5% of garments need rework, you are effectively paying $1.50 for work that now needs to be partially redone. The true cost of poor quality is not just the rework — it is the full production cost of garments that did not come out right the first time.

How We Got from 8% to 2.7% — Week by Week

Here is the actual timeline of our quality improvement. No cherry-picking. No exaggeration. These are the numbers from our daily inspection reports.

The biggest drop happened in Week 1, just from moving to inline inspection. This is the lowest-effort, highest-impact change any factory can make. You do not need software. You do not need new equipment. You need one person with a keen eye roaming the line and giving immediate feedback.

Week 2's machine audit was the second biggest drop. Again, zero capital investment — the mechanic was already on payroll. We just changed how he spent his time, from reactive repairs to proactive audits.

Weeks 3 and 4 required data — tracking defects by operator and operation, which is where a digital system helps significantly. You can do it on paper, but the pattern recognition is much slower.

Using Data to Drive Quality

Once you have operator-level defect tracking, patterns emerge that are invisible to the human eye. Here are the most useful dimensions to analyze:

By Operator

Some operators consistently produce higher defects on specific operations. This is not about blame — it is about fit. An operator who produces 6% defects on single needle but 0.5% on overlock should be on overlock. The data makes the case without any argument.

By Operation

Certain operations are inherently defect-prone. Collar attachment, armhole setting, and any operation involving curves tend to have higher DHU. These operations need more experienced operators and closer inline inspection. Knowing which operations are your quality risks lets you allocate inspection time where it matters.

By Machine

A machine that produces consistently higher defects than other identical machines on the same operation has a mechanical problem. We found one overlock machine that produced 3x the skip stitches of the machine next to it. Same operator skill level, same fabric, same thread. The machine had a worn looper that nobody had noticed because skip stitches were attributed to the operator.

By Time of Day

This was the most surprising finding. Our defect rate was 40% higher between 2 PM and 4 PM compared to the morning. Post-lunch fatigue is real. Friday afternoon was our worst quality window of the week. Once we knew this, we scheduled inline inspection more heavily during afternoon hours and moved the most critical operations to morning shifts.

By Day of Week

Monday mornings had higher defect rates than the rest of the week, especially in the first hour. Machines had been sitting idle over the weekend, thread tension drifted, and operators needed time to get back into rhythm. We added a 15-minute warm-up period on Mondays where operators sew scrap fabric to calibrate machine settings and hand coordination before starting on production pieces.

Common Mistakes That Keep DHU High

In working with other factories, I see the same quality mistakes repeated everywhere:

• Inspecting only at end-line: By the time you find the defect, hundreds of pieces may already be affected. Inline inspection is non-negotiable if you want DHU under 3%.
• Blaming operators for machine problems: Always check the machine before blaming the person. At least 40% of sewing defects originate from machine settings, not operator skill.
• General quality training: "Be more careful" is not actionable advice. Train on the specific defect, at the specific machine, with the specific fabric. Five minutes of targeted training beats an hour of lectures.
• Accepting fabric without inspection: Every defective roll that enters your cutting room will exit as defective garments. Fifteen minutes of fabric inspection per roll saves hours of rework.
• Not tracking quality data: You cannot fix what you cannot see. If you do not know which operator, which operation, and which time of day produces the most defects, you are guessing. And guessing does not get you below 3%.
• Treating rework as normal: Some factories budget for rework as if it is an inevitable cost of production. It is not. Rework is a failure of the quality system. Every garment that goes to rework is a garment that should have been done right the first time.

Quality Is Free — Rework Is Not

Philip Crosby wrote a book called "Quality Is Free" in 1979. The title sounds like marketing, but the math is solid. Preventing a defect costs almost nothing — a properly maintained machine, an inline check, a 5-minute training session. Fixing a defect after it has been sewn costs 3-5x the original operation. Fixing a defect after it has been packed and shipped costs 10-50x in penalties, air freight, and lost buyer confidence.

Every dollar you invest in prevention returns $5-20 in avoided rework. The factory that spends $200/month on inline inspection and machine maintenance saves $3,000-5,000/month in rework, rejected shipments, and overtime to meet deadlines that slipped because of quality problems.

Quality is not a cost center. It is the cheapest way to increase your effective output. Reducing rejection from 8% to 3% on 2,000 daily pieces means 100 more saleable garments per day. That is the same as adding 100 garments of capacity without hiring a single person or buying a single machine.

We went from 8.2% to 2.7% in 30 days. Not because we hired better operators or bought better machines. Because we started seeing defects where they happen, when they happen, and we fixed the system instead of blaming the people. If your rejection rate is above 3%, the same approach will work for you. The defects are not random. They have patterns. Find the patterns, and the fixes are usually simple.

Santosh Rijal is the founder of Scan ERP, a garment manufacturing ERP system designed for factory floor operations. He works directly with sewing lines, cutting rooms, and production supervisors across Nepal's garment manufacturing sector.