Smart IE: Digital Industrial Engineering in a Garment Factory
Quick answer. “Smart IE” is industrial engineering run on data captured on the sewing floor as work happens, rather than on occasional stopwatch studies. One example is an observed SMV taken from scan timestamps:
Observed SMV = 60 ÷ observed pieces per hour
Illustrative example: an operation has a standard SMV of 1.10 minutes. Scans show operators doing it at 48 pieces an hour. Observed SMV = 60 ÷ 48 = 1.25 minutes, 0.15 minutes above the standard. The gap is a question for the IE, not an answer: method, skill, waiting or a wrong standard.
Traditional garment IE runs on two kinds of data: a time study done when a style is set up, and output counted by hand during the day. Both are slow. A time study is a snapshot of one operator on one day; a tally sheet arrives hours after the problem it records. This guide looks at what changes when the same IE questions are answered from data the floor produces continuously, what the published research actually shows, and what that data cannot do. The formulas themselves do not change; they are all collected in our IE formula reference.
What is digital industrial engineering in the garment industry?
Digital industrial engineering means the IE works from production data captured on the floor as the work happens, by scans, sensors or tablets, instead of from occasional stopwatch studies and hand-written tallies. The formulas stay the same. What changes is how fresh, complete and per-operator the inputs are, and how quickly a problem shows up.
Textile Learner's piece on smart factories describes industrial engineering as "the field of study that is concerned with the design and optimization of complex systems", and lists the IE's work in a smart factory as time and motion study, method improvement, line balancing, capacity planning and workflow design, with the added job of making sure digital tools fit the way the floor actually works (Elma, Textile Learner, 2025). That is the useful framing: the data is new, the discipline is not.
What changes when IE data comes from the floor
The problem with manual data was described plainly in an Apparel Resources (StitchWorld) piece back in 2011: "Manual data collection is also laborious and error prone." The same piece estimated that collecting individual production figures from a 30-machine sewing line may take 15 to 20 minutes, and argued that collecting performance data directly from the operator turns it into information in real time (Apparel Resources, 2011). Here is what that means for each part of the IE's job:
| IE question | With stopwatch and tally sheets | With real-time floor data |
|---|---|---|
| How long does this operation take? | Standard SMV from a time study of a few cycles, rarely repeated | Observed cycle time from bundle scan timestamps, compared with the standard SMV |
| Who should do this work? | Supervisor's memory of who is good at what | Skill records built from each operator's speed and defect history per operation |
| Where is the bottleneck? | Walk the line and look for piles of bundles | Work piling up at one operation while the next starves, visible as it builds |
| Who is idle? | Seen only when someone walks past | Operators with no active work, matched against queued work on their machine type |
| What is stuck? | Found at lot close, when a bundle is missing | Bundles that have not moved for longer than expected, flagged during the day |
| Is quality slipping? | DHU totalled at day end | Defects recorded per check, per operator and operation, through the day |
Standard SMV vs observed cycle time
Prabir Jana made the IE case for observed times in 2010: "Cycle time is more important than standard time as it gives an accurate indication of what can be achieved", and "for balancing purposes, it is much better to use achieved cycle times than standard times" (Jana, Apparel Resources, 2010). The catch has always been collecting achieved cycle times for every operation without a second person with a stopwatch. When each bundle is scanned at the start and end of an operation, the timestamps give that cycle time as a by-product.
What the research shows
Three studies are worth reading in full, because they show both the gains and how they were measured:
- Real-time mobile line balancing, Thailand. Siriruk and Pengsiri built a mobile and web app, tested at a garment factory in Thailand, that replaced stopwatch timing and spreadsheet line-balancing with digital time recording and an automatic Yamazumi chart. The abstract reports line-balancing work falling "from four hours to under ten minutes"; the results section describes the old method as "over 3 hours" and the app as "less than an hour". After bottlenecks were removed, production efficiency rose from 70.4% to 76.3% in the test case (IEOM Asia Pacific Conference, Tokyo, 2024). Note that the app still records times per task; it speeds up the study rather than removing it.
- Sensors and RFID on the sewing line, Turkey. Koç and Eryürük combined real-time data from process monitoring devices on sewing machines, with operators scanning RFID cards, with a weighted line-balancing algorithm that assigns operators to stations. In their case study, line efficiency rose from 79.68% to 88.31% and production per operator by 10% (Fibres & Textiles in Eastern Europe, 2025).
- Low-cost end-of-line capture, Li & Fung suppliers. Kevin Hsu's MIT thesis developed "a cost-effective, digital tool to capture quality and production data at the end of a production line" for factories that captured data manually, and tested it in three factories in India and Malaysia. The abstract describes expected benefits, such as real-time analytics and faster root-cause correction, rather than measured results (MIT DSpace, 2019).
These are single case studies in particular factories, not industry benchmarks. What they share is the mechanism: fresher data about where time is lost, followed by an engineer acting on it.
Does real-time scan data replace the time study?
No. Scan timestamps show how long work actually took, waiting and interruptions included, which is an observed cycle time. A standard time needs a defined method, a performance rating and allowances, which only a time study sets. Scan data is best used to check standards against reality and to see where the line is losing time.
An observed SMV that is well above the standard can mean four different things: a slow method, an operator still learning, a line that kept her waiting, or a standard that was wrong to begin with. Scan data narrows the search, but deciding which one it is still means going to the machine. The rating and allowance method is in our SMV and SAM calculation guide.
What Scan ERP does for IE today
Scan ERP is built around QR bundle scans, so most of what it offers an IE comes from those scans and from biometric attendance. In IE terms:
- Standard minutes on every work item. SMV and piece rate per operation are set in the style template and copied to every work item; standard minutes for a lot = garments × SMV. An operation dictionary suggests SMV and price ranges by operation name.
- Template critical path. When a style template is built, it works out each operation's earliest finish (earliest start + SMV), so the longest chain of operations is visible before the style reaches the floor.
- Observed SMV from scans. The lot-close report shows a measured cycle time per operation from scan timestamps: observed minutes per piece = 60 ÷ a robust 75th-percentile speed (pieces per hour, with the fastest 10% of readings dropped) ÷ pieces per garment. The software itself treats scan timing as approximate, and so should you.
- Worked minutes per operator and operation from scan start and end times, falling back to the gap since the last scan (capped at 30 minutes) or to attendance check-in.
- Piece-rate suggestions (shadow mode). A suggested fair rate from the 75th-percentile speed, capped at ±10% per cycle. It only suggests; it never changes a price by itself.
- Skill matrix. A skill score from 0 to 100 per operator per operation, from speed against the team, defect rate and consistency, with tiers from New and Learning to Competent, Skilled and Expert, plus a “Free now” list of operators with no active work and operations they could learn.
- Smart work suggestion. When an operator taps “I am free”, ready work is ranked by same-operation and colour continuity, starved bundles, lot close-out, fairness and downstream readiness, keeping operators on their main machine and steering risky operations away from operators with high defect rates.
- Line-flow gates and bottlenecks. It flags an operation piling up work while the operations after it starve, with the backlog in person-days; lists the longest machine-type queues next to logged-in idle operators who run that machine; and detects finishing imbalance, such as ironing out-producing folding.
- Stuck work. Stuck, behind and stalled bundles are flagged, and work whose dependencies are complete is released automatically. See WIP tracking.
- Pace and output signals. On the floor TV: each machine type's share of today's standard minutes and a lot completion estimate in days (remaining SMV minutes ÷ the 7-day average of daily SMV minutes). Daily output is compared with a fixed daily target the factory sets, using the expected pace for the time of day (green, amber, red), with an end-of-day projection, and an hourly grid compares pieces per hour with the day's own average and with yesterday.
- Idle signals. An idle-operator list that leaves out people who have punched out, voice announcements for idle iron, fold and QC stations, and finishing stall alerts.
- Quality. Roving QC traffic-light checks (green, yellow, red) and DHU in the QC scorecard and daily digest; see garment QC software.
- Time and pay. Piece-rate payroll from scans, and earnings per hour per operator using worked minutes from ZKTeco biometric attendance.
- WhatsApp reports. Bottleneck alerts at 11 am and 3 pm, end-of-day output against the 5-day average, and smart insights through the day.
What it does not do (yet)
- No stopwatch time-study module, and no rating and allowance workflow.
- No GSD or other predetermined motion time system; SMV is entered by the factory or taken from the dictionary ranges.
- No SMV-based operator or line efficiency report (earned minutes ÷ minutes paid for). Our line efficiency guide shows how to calculate it.
- No line-balancing tool: no takt or pitch diagram, no operator-to-station allocation, no balance-loss figure.
- No hourly target per line or operation, and no target calculated from SMV and manpower; the daily target is a number the factory sets.
- No capacity planning, line loading or Gantt chart.
- No logging of non-productive time or downtime reasons.
- No “line” as a unit in the software: work is grouped by lot, machine type and operator.
Where does a human IE still matter?
In three places data cannot cover: improving the method of an operation, setting the standard time a new style is costed and paid on, and deciding how to rebalance a line, which means weighing operator skills, machines and the order at hand. Real-time data tells the IE where to look; the IE still decides what to change.
- Method improvement. A folder, a guide or a better hand position changes the cycle time. Scans will show the improvement afterwards; they cannot invent it.
- Setting standards. Costing and piece rates need a standard that both buyer and operator can trust. That comes from a rated time study or an agreed SMV, with observed times as a cross-check.
- Balancing decisions. Knowing that one operation is starving the next is half the job. Choosing whom to move, onto which machine, and for how long, is the other half. The arithmetic is in our line balancing guide; the judgement is the IE's.
What an IE does all day, and which parts of it software can take over, is covered in our guide to the industrial engineer in a garment factory.
Sources
- Pavee Siriruk and Sirintorn Pengsiri, "Real-Time Mobile Labor Sewing Line Balancing For Garment Production Factory", 5th Asia Pacific IEOM Conference, Tokyo, 2024. DOI 10.46254/AP05.20240222.
- Bülent Koç and Selin Hanife Eryürük, "Optimizing Sewing Line Balancing in Apparel Manufacturing through Digitalization", Fibres & Textiles in Eastern Europe 33(1), 2025. DOI 10.2478/ftee-2025-0002.
- Kevin Ta-Zhi Hsu, "Using factory-level digital tools to improve quality and productivity in garment factories", MIT, 2019.
- Prabir Jana, "Understanding Time's measures to achieve better production", Apparel Resources, 2010.
- Apparel Resources (StitchWorld), "Real time shop floor data collection critical to productivity", 2011.
- Shafiun Nahar Elma, "Smart Factories and IE: Shaping the Future of Garment Manufacturing", Textile Learner, 2025.
See Your Floor's Real Cycle Times
Scan ERP turns QR bundle scans into observed SMVs, a skill matrix, bottleneck and idle alerts, and piece-rate pay, with worked minutes from ZKTeco punches. Ask the founder on WhatsApp and see it on a real line.
Ask on WhatsApp Request a Free Demo