Picture-based work instruction cards
Quality

Picture-based work instruction cards

Financial impact

$8,400USD/ yr EBITDA

Operational impact

~6fewer reworks / day
high confidence

The math

Annual defects at baseline11,200

112000 garments × 10.0% defect rate

Rework cost avoided$8,400/ yr

defects × 15% reduction × $5/defect

Capex

$500

Payback

1 months

Difficulty

2 / 5

Time to implement

~4 weeks

Drop in defects
%
Cost to fix one defect
$
Printing and lamination
$

Also uses annual garment volume 112,000, in-line defect rate 10%. Edit on the dashboard.

Description

Right now nothing at the sewing stations tells a worker the correct way to do their task. No photos, no written standard, no target; the assessor found none anywhere on the floor. Each worker does the job from memory and habit, so the same operation comes out a little differently from person to person and shift to shift. That variation is a big source of the faults that then have to be unpicked and re-sewn, which burns labor and slows the line.

Evidence

  • [00:00:33] Bare machine heads and tables, no instruction card at any station
  • [00:01:00] No production targets or standards posted anywhere on the floor
  • [00:01:00, 00:01:08] Small Indian flag on the thread stands, a South Asian workforce that needs translated cards
  • [00:01:15, 00:02:50, 00:04:11] Parts moving between stations in brown shopping bags, the reason a card must travel with the bundle

Quotes

No standardized work instruction sheets… were visible.

Assessor

Deep dive

Expand for detail

Load-bearing, card breaks if wrong

Most of the faults come from people doing the job differently, not from bad machines or fabric.

A written card only removes faults caused by method: one worker sets a different stitch length, another lines up a collar differently, so the results vary. If the real cause is a worn machine, a dull needle, bad fabric, or dim light, a laminated card changes nothing. This one assumption is what the whole $8,400 rests on. What it looks like on the floor if this is wrong: you post the cards, workers follow them, and the defect count does not budge, because the faults were coming from the equipment all along. The cheap insurance is to pull the repair log first and check whether the top faults are the method kind or the machine kind; the machine kind belong to the daily-maintenance card, not this one.

The card travels with the fabric bundle, not stuck to the machine.

This factory runs many small orders at once and moves parts between stations in brown paper shopping bags (you can see them at [00:01:15], [00:02:50], [00:04:11]). A worker sews shirt collars one hour and jacket sleeves the next. If a single "shirt collar" card is zip-tied to the machine, it is dead wrong the moment a bag of jacket parts arrives. The fix is that the card rides inside a clear plastic sleeve on the bundle itself and gets clipped up only for that job. What it looks like on the floor if this is wrong: a worker sewing jacket sleeves while a shirt-collar card hangs in front of them, and within a week everyone treats every card as noise.

Supervisors actually point to the card and correct people on their floor walks.

A laminated sheet does not change habits; a supervisor does. The floor supervisor (yellow polo at [00:00:48]) has to point at the card and correct anyone who drifts from it, every day, or workers slide back to old habits within a week. What it looks like on the floor if this is wrong: the cards are up, but by week six they are buried under tools, water bottles, or draped fabric [00:03:00], and the supervisor walks past without a word.

Supportive, card weakens but survives

The cards are photo-led and in the workers' own languages.

The floor is almost entirely South Asian labor; there is even a small Indian flag mounted on the thread stands [00:01:00, 00:01:08]. An English-only, wordy card is invisible wallpaper to a worker who reads Hindi, Urdu, or Bengali. Cards have to be mostly photos, with 5 to 10 words at most, translated. What it looks like on the floor if this is wrong: workers glance at the card once, cannot use it, and go straight back to asking the person next to them.

One person can document the 'right way' without stopping production.

The factory has no industrial engineer (the specialist who normally writes work instructions), so the job falls to the master sample-maker or the floor supervisor: photograph the correct method at each station, one station at a time. There are 170-plus distinct operations across shirts, trousers, and jackets, so you cannot do them all; you document the busiest handful first. What it looks like on the floor if this is wrong: the supervisor is pulled off the documentation every hour to firefight, and three weeks later only five stations have cards.

The card does not travel with the fabric, so it is wrong the moment the job changes

Risk

This is a high-mix, build-to-order floor: a worker switches from shirts to jackets to trousers through the day, and parts move around in brown shopping bags. If the card is fixed to the machine instead of riding with the bundle, the worker is looking at instructions for a different product than the one in their hands.

Early Warning

A mid-day walk: pick five stations at random. If the fabric in the worker's hands does not match the product pictured on the card in front of them, the system has already failed.

Mitigation

Put the card in a clear sleeve on the bundle's shopping bag or job ticket, so it physically travels with the work. The worker clips it up for that job, and it leaves with the bundle. Train the material handlers that swapping the card is part of dropping off the bundle.

Stake

The whole $8,400. Worse, workers learn to ignore every card, which poisons any future attempt at written standards.

The cards go stale because the product changes every order

Risk

Because the factory only makes what is already ordered, this month's jacket is not last month's polo shirt. A card photographed for an old style is not just useless, it teaches the wrong method and destroys trust in the whole system.

Early Warning

Cards on the wall show a garment nobody is making this week, and workers have stopped looking at them.

Mitigation

Version and date every card, name the supervisor as owner, and refresh the cards as part of each new order's setup step, not as a separate task. Focus the cards on the highest-volume, longest-running styles, where they stay valid longest.

Stake

The $8,400 erodes toward zero as the covered operations drift out of date, and the lost trust makes the next attempt harder.

The faults were machine or fabric-driven, so the cards change nothing

Risk

Cards only fix faults that come from people doing the job differently. If the real causes are worn machines, dull needles, bad fabric, or poor light, the defect count will not move no matter how good the cards are.

Early Warning

Workers are clearly following the cards, but the defect tally on those operations is flat after two order cycles.

Mitigation

Pull the repair log before you start and confirm the top faults are the method kind. Send the machine-kind faults to the daily machine-cleaning and maintenance card (rank 5) instead, where they actually get fixed.

Stake

The full $8,400 fails to appear, and the effort is spent on the wrong problem.

No one audits, so it quietly backslides within weeks

Risk

The supervisor is busy firefighting the pressing bottleneck and stops checking the cards. Within a few weeks the cards are buried under tools and fabric and everyone reverts to old habits.

Early Warning

By about week six, cards are draped over or hidden under tools, water bottles, and fabric [00:03:00], and the supervisor walks past without correcting anyone.

Mitigation

The factory manager audits the supervisor, not just the workers: add "are the correct cards up and in use" to a daily 5-minute manager walk, and have the owner review it weekly.

Stake

Defect rates drift back to the 10% starting point within a month, and the whole gain is lost.

A. What one good instruction card looks like

An instruction card (the garment industry calls it a "visual SOP," where SOP just means the single agreed best way to do one task, written down) is one laminated sheet per operation, mostly pictures. It shows the correct method side by side with the common fault, so a worker who reads little English can follow it at a glance.

================================================================
STATION: Fusing press (Lastar)   PRODUCT: Standard shirt   VERSION: 1.0
================================================================
        CORRECT METHOD (green tick)        COMMON FAULTS (red cross)

[Photo: wide shot, temperature dial]     [Photo: scorched, peeling lining]
Temp 160C. Speed 4. (Hindi / Urdu)       WRONG TEMP OR SPEED

[Photo: glue side DOWN]                  [Photo: glue smeared on the belt]
Glue side DOWN. (Hindi / Urdu)           GLUE SIDE UP

[Photo: collar caught flat on exit]      [Photo: collar creased]
Stack FLAT. Do not fold. (Hindi / Urdu)  FOLDED WHILE HOT
================================================================
Supervisor sign-off: [____]   Next review: [tie to the next order's setup]

(The fusing press is the heated machine that glues the stiffening layer into collars and cuffs. Get its temperature wrong and the lining scorches or peels, which is why it is a good first card.)

What it costs in Dubai. About 60 cards, color-printed on A4 and heavily laminated, run roughly $3 each, so about $180. Add $150 to $200 for the clips, zip-ties, clear bundle sleeves, and a cheap phone or tablet to take the photos. Print shops in Deira and Al Quoz turn these around same-day, and the total lands near the $500 budgeted. The real cost is not the printing, it is the supervisor's time to photograph and document the right method for each operation.

Languages. The floor is almost entirely expatriate labor from India, Pakistan, and Bangladesh, so the pictures carry about 80% of the meaning and the text is just a label. Print each caption in English (for managers and any auditor) plus Hindi, Urdu, and Bengali, kept to 5 to 10 words with no complex grammar ("Stack FLAT. Do not fold."). Confirm the exact wording with the operators themselves so it matches how they actually talk.

B. Why cheap laminated cards, not screens, and why the card must travel

Two fair questions: why not put a tablet or screen at each station, and why not just bolt one card to each machine?

  • Cards, not screens. A screen at every one of 56-plus stations means thousands of dollars of hardware, plus power, network, breakage, and one more thing to maintain on a floor that already skips machine cleaning. Laminated cards cost a few dollars, need no power, shrug off oil and steam, and can be reprinted in an afternoon. Screens only earn their keep in high-wage, high-tech plants; here they are a liability.
  • Photos, not paragraphs. The floor is almost all South Asian labor (there is even an Indian flag on the thread stands at [00:01:00]). A wordy English card is invisible. Photos with a green tick for right and a red cross for wrong carry the meaning; the few words are just labels.
  • The card must ride with the work. This is the make-or-break detail. The factory runs many small orders at once and moves parts in brown shopping bags [00:01:15, 00:02:50, 00:04:11]. A worker sews shirt collars one hour and jacket sleeves the next, so a card fixed to the machine is wrong the instant the job changes. The card lives in a clear sleeve on the bundle and travels with it; the worker clips it up only for that job. Get this one thing wrong and the whole program dies in a week.

C. Prove it is worth doing before you print a single card

The entire $8,400 rests on one thing: that the faults come from people doing the job differently, which a written card can fix. Faults from worn machines or bad fabric will not move. Prove it first.

Pull these and check

  • Repair and inspection log, last 30 to 60 days (floor supervisor or QC room): sort the faults by cause. Are the top ones the method kind (misaligned collar, wrong seam) or the machine kind (skipped stitches, scorch)? Only the method kind are yours to fix.
  • The product mix, top styles by volume (sales or factory manager): the "80/20" rule, meaning the small share of styles that make up most of the output. There are 170-plus operations across shirts, trousers, and jackets; you document only the 10 to 15 that cover about 80% of the volume.

The go / no-go rule

  • Go if the top faults are mostly the method kind and a handful of styles cover most of the volume: build cards for those operations first.
  • No-go if the top faults are machine or fabric-driven: stop, and put the money into the daily machine-cleaning and maintenance card (rank 5), which is where those faults actually get fixed.

What we did not have in the corpus and would request from the client on day one. The aggregated gaps across all 15 cards become the engagement-letter ask sheet.

  1. 01

    Repair / rework log (last 30 to 60 days)

    Owner

    Floor supervisor or QC room inspector

    To check the 10% fault-rate assumption and, more important, to see whether the top faults are the method kind (fixable by a card) or the machine kind (not). This one dataset decides whether the card is worth building and which operations to document first.

  2. 02

    Payroll plus a stopwatch sample of repair time

    Owner

    HR / Accounts, with the floor supervisor

    To ground the $5 cost to fix a fault, which is the least certain number on the card. Pure sewing labor to unpick and re-sew is well under a dollar, so we need to confirm the handling, re-checking, and line disruption that make up the rest.

  3. 03

    Workforce language roster

    Owner

    HR or the accountant

    To confirm exactly which languages to print (Hindi, Urdu, Bengali, or others), so every worker on the floor can actually read the cards. English-only cards will be ignored.

  4. 04

    Top styles by volume (the product mix)

    Owner

    Sales or factory manager

    To scope the work. If a few basic shirt and trouser styles drive most of the output, we build cards for those first instead of trying to document all 170-plus operations.

  5. 05

    Production counts

    Owner

    Factory manager / Accounts

    To confirm the roughly 112,000 garments a year that every number on this card is built on.