Research libraryUtilisation Decides
Chapter 09 of 14 · Part III · Draft web edition
Balanced work needs controlled release. A line shortens delivery time when it admits work that downstream operations can use, makes queues visible and protects good output at its constraint.
The cloth is woven from this chapter's 5 exhibits: each weft is one exhibit's own data, normalised; each warp group is a section. The probe reads the exhibit's own value where it crosses. Procedural visualisation, not a measurement.
Read-out
GFT discussion model, not measured scores. Axes run from low to high on an arbitrary 0 to 10 scale. Push is an authorisation policy; bundle and direct handover are transfer arrangements and may coexist with different release policies. Positions indicate questions for a trial, not rankings or recommended limits.
Unit: illustrative ordinal discussion scale. Status: illustrative. Sources: S22, S24. Illustrative figures are teaching examples, not measured factory results.
Overture · 1 min
Lean flow requires a release rule, an explicit work-in-progress boundary and a response when that boundary fills. Balancing establishes a feasible distribution of work; controlled release prevents that distribution being overwhelmed by material admitted faster than the line can finish it. GFT recommends managing accepted garments leaving the line together with the quantity and age of work inside it. An operator's activity is useful when it advances authorised work towards completion. Producing components without a downstream need can increase activity while leaving delivery unchanged. This chapter sets out a proposed control design, distinguishes evidence from assumptions and explains how a factory can test the design before extending it.
9.1
Related · chapter 0111 paragraphs step it
Read-out
GFT proposed review sequence: check time losses, labour content and flow, then choose an action. OEE, earned minutes and stable-system WIP relationships remain distinct measures. Arrows indicate review order, not material flow or a combined score. This is not an implemented platform.
Unit: illustrative measurement design. Status: illustrative. Sources: S04, S13, S22. Illustrative figures are teaching examples, not measured factory results.
Read this exhibit in its chapter ↗The Lean Enterprise Institute defines pull through a downstream request specifying what is needed, how much, where and when. Upstream production follows that request. Its supermarket approach replenishes withdrawals from a controlled store, whereas sequential pull uses a planned product sequence where stocking every item would be impractical. These are different ways of authorising work; a kanban card is a carrier of the instruction.[S24]
GFT proposes defining a release decision before choosing its technology. The instruction should identify the garment family, operation, permitted quantity, destination and material status. It should also name the person responsible for rejecting an incomplete instruction. A brightly coloured card that travels with every bundle irrespective of downstream capacity changes the label on production rather than its governing decision. An electronic message has the same limitation if the receiving process cannot refuse work when its agreed space is full.
For a proposed sewing pilot, the control boundary should start when a complete, approved kit enters the line and end when the garment passes the agreed exit inspection. The kit definition must specify matching panels and required trims, without equating a loose panel with a garment. Record partially completed kits separately until their relationship to a complete garment is known. Otherwise, the release count can appear conservative while incomplete pieces occupy machines and floor space inside the boundary.
A release permission should be consumed by a stated event and returned by another stated event. GFT recommends consuming it when the kit crosses the entry point and returning it when an accepted garment leaves. Scrapped work needs an authorised closure event, with the material discrepancy recorded. Rework retains its identity and remains visible within the boundary. These proposed rules prevent a rejected garment from generating fresh admission merely because it has been moved to a repair table.
The supervisor should also have a documented refusal rule. If the permitted queue is full, the default action is to stop further admission at that interface and investigate. A planner may authorise a temporary exception, but the exception should state the extra quantity, reason, expiry and expected recovery. Treating every urgent order as an unrestricted exception would make the nominal limit meaningless. A visible exception record lets management distinguish an unsuitable limit from repeated avoidance of a suitable one.
Pull should not be confused with an instruction to run down all upstream material stocks. A sewing-line admission limit answers how much authorised work may occupy the sewing system. The purchasing and fabric policies answer different questions about replenishment and supply risk. GFT recommends maintaining separate records for these boundaries so that a line pilot cannot claim success by moving its queue into the cutting room. Inventory policy across the wider supply chain belongs in the paper addressing that subject; here the question is disciplined admission to production.
The practical test is whether anyone can point to the permission for a garment now inside the line. If the answer depends on remembering a conversation, the design needs a clearer record. If permission exists but no usable material accompanies it, investigate preparation. If material exists but the permitted space is occupied, investigate flow. Keeping these questions separate makes the response specific without attributing every delay to a machine or to an operator.
GFT recommends a release ledger whose record can survive a shift handover. Its minimum fields should describe the kit identity, product, quantity, admission time, current location, disposition and permission status. Corrections should preserve the original event and explain why it changed. A counted garment without a matching event is an exception to investigate, not a reason to backfill a convenient admission time. This proposed discipline gives the next supervisor a recoverable operating state.
A lost card should trigger reconciliation before replacement. Check the physical kit, its last recorded location and any completed exit. Cancelling the original permission and issuing its replacement should be linked in the record. Otherwise, both may later circulate as apparently valid instructions. The same rule applies to an electronic instruction resent after a connection failure: a repeated message should refer to the existing permission rather than create additional authorised quantity.
During a power interruption, GFT proposes freezing new admission unless the approved contingency method can preserve identity, safety and the permitted inventory boundary. At restart, reconcile work still on machines, work under inspection and work awaiting repair before releasing replacements. A recovered connection should not automatically replay every queued instruction as fresh production. The proposed control objective is continuity of authorisation across the interruption, rather than continuity of machine activity at any cost.
9.2
Related · chapter 079 paragraphs step it
Read-out
Illustrative GFT six-gate validation sequence for accepting a proposed sewing-line balance. Precedence and resource constraints follow S15; pitch-time bounds follow S16; operator skill-level matching follows S17. Every candidate allocation must pass a physical line trial before becoming standard work.
Unit: . Status: illustrative. Sources: S15, S16, S17. Illustrative figures are teaching examples, not measured factory results.
Read this exhibit in its chapter ↗MIT's lean glossary describes single-piece flow as work moving immediately from an operation to the next, with no intervening buffer between workers. This concerns the transfer arrangement, not an empty production system: work can still be undergoing processing at each operation. For this chapter, a progressive bundle means that a group waits to move together, while a small transfer bundle is a deliberately bounded intermediate design.[S22]
GFT recommends comparing those arrangements by the response they require when work cannot advance. In a strict direct handover, a downstream refusal should become immediately visible upstream. In a bounded transfer queue, the same refusal first consumes the agreed waiting space. With unrestricted bundles, material can continue accumulating without forcing an explicit decision. These are proposed diagnostic questions for a line walk, rather than a claim that every observed bundle system behaves identically.
Choose the first pilot boundary where adjacent operations have compatible work and an available response to abnormalities. Require the receiving operator to confirm identity and condition at handover. Agree who answers a stop signal and what work is safe while the issue is resolved. Removing waiting space before the team can answer that signal would test endurance rather than the proposed control method. The trial should expose problems in a way that the team can act on.
Small transfer bundles can be a reasoned starting point where direct handover is not yet workable. Define the bundle by its maximum quantity and identity, and define the receiving queue separately. The production lot, transfer bundle and release allowance need not be equal. A style may remain the planned production lot while its components move in smaller groups. That distinction lets a pilot alter transfer behaviour without silently rewriting purchasing quantities or the entire style schedule.
Quality containment needs its own rule. GFT recommends recording the last confirmed conforming handover when a seam fault is found, then tracing the potentially affected work using its identities. Do not send doubtful pieces forward to preserve apparent flow. Mark their status, retain them in the inventory reconciliation and give their disposition an owner. The useful question is whether a defect is found while the response can still contain it, rather than whether an inspection point has been removed.
The proposed design comparison in Exhibit 40 uses invented ordinal positions to structure discussion. Its horizontal direction represents increasing stability of the local process. Its vertical direction represents increasing product variety. The positions are illustrative judgements, not measurements or an optimisation. Push describes an authorisation policy; bundles and direct handover describe transfer arrangements. They are shown together to expose design choices, not to imply mutually exclusive manufacturing categories.
Before selecting a point on this illustrative map, GFT recommends asking the team to demonstrate the assumed conditions. Can the chosen sequence be maintained? Can matching components be delivered together? Can a refusal be resolved without bypassing inspection? Record each answer as observed, untested or failed. A transfer design should proceed on the strength of those observations, with an agreed fallback, rather than because its name appears further towards the stable end of a diagram.
A proposed handover test should also cover component separation. If a garment's panels travel along different preparatory routes, direct transfer of an isolated panel does not prove that the complete garment can advance. Require matching identities at the joining operation and record missing counterparts as a preparation exception. Keep the transfer rule distinct from the completion count, so that a fast stream of available panels cannot hide an older queue of incomplete garments.
GFT recommends involving operators in choosing the physical holding method. A permitted queue should have a defined location and safe presentation, without requiring excessive reach or obstructing access. Demonstrate how work is returned when a handover is refused. If the agreed space cannot hold the permitted work safely, revise the allowance or the arrangement before the trial. The inventory rule and the physical method should describe the same feasible operation.
9.3
Assume stable flow
Assume stable flow. Starting point. Leads to Mean WIP 20 or 10. · illustrative
Little’s Law relates average WIP to average throughput rate multiplied by average time inside the same system.
Mean WIP 20 or 10
Mean WIP 20 or 10. From Assume stable flow. Leads to Divide by 5/hour. · illustrative
Divide by 5/hour
Divide by 5/hour. From Mean WIP 20 or 10. Leads to Mean time 4 or 2 hours. · illustrative
Mean time 4 or 2 hours
Mean time 4 or 2 hours. From Divide by 5/hour. End point. · illustrative
The method behind it: Worked example: identical throughput can coexist with different WIP.
Illustrative stable boundary with no scrap, rework or bypass. Assume accepted throughput is 5 garments/hour. Little's Law gives mean time = mean WIP / throughput: 20/5 = 4 hours; 10/5 = 2 hours. Assumed values demonstrate the identity and do not establish feasible inventory limits.
Unit: illustrative garment equivalents and elapsed hours. Status: illustrative. Sources: S22. Illustrative figures are teaching examples, not measured factory results.
9.3 · the bodyExhibit 41 above; its figures are illustrative where marked.
Little's Law relates average WIP to average throughput rate multiplied by average time inside the same system. MIT's glossary states the long-run stability condition explicitly. Here, throughput time means elapsed residence from kit admission to exit, including waiting and repair. It must not be confused with the shorter repeat time of a workstation. The relationship concerns compatible averages, not a promise for an individual garment.[S22]
For a proposed pilot, GFT recommends drawing the boundary on the floor plan and on the collection sheet. Count work waiting, being processed and held for repair or disposition inside it. Use an elapsed-time clock for residence time and an output rate expressed on that same clock. An overnight stop remains part of calendar residence if the garment stays inside. Alternatively, a declared operating-time clock can answer a different question, provided all terms use it consistently.
The time-weighted inventory average deserves particular attention. A photograph of an unusually tidy queue at the end of a shift is not an average. GFT proposes logging each entry and exit where practical, then integrating the inventory level over the observation interval. Where manual collection requires snapshots, spread them across the interval and report the approximation. Keep opening and closing inventory visible, and explain any transfers whose time or identity could not be reconciled.
The illustrative relationship is made explicit in Exhibit 41. The diagram is a calculation aid with assumed inputs. It does not size the required stock ahead of a constraint or establish the time needed for replenishment. Those decisions require evidence about actual interruptions and recovery.
GFT recommends treating a discrepancy between the measured averages as a reconciliation task. Check whether the inventory counts include repair while the output count excludes its eventual exit. Check whether work crossed an unrecorded boundary, whether time bases differ, or whether the observation covered a deliberate inventory drawdown. Do not adjust the inventory count solely to force agreement. A transparent explanation of a transient period is more useful than a precise-looking time figure built from incompatible data.
For a line losing work to scrap, the proposed collection sheet should retain accepted output and scrap exits separately. Physical flow conservation must account for every exit; the simple good-output-only example above deliberately excludes that complication. Report cohort residence for accepted garments separately if that is the customer-service question. Do not divide all inventory indiscriminately by good exits and describe the result as a universal residence measure. The boundary and population should remain inspectable by the reader.
A proposed observation plan should retain unfinished work at the end of the measurement window. Completed garments alone can present a favourable picture if difficult jobs remain inside. Record the age and status of those open jobs, then follow them to disposition where practical. Describe whether the residence measure covers an admission cohort, an exit cohort or a continuing steady period. These are different views, and GFT recommends naming the chosen view before calculating its average.
The same discipline applies to unusually old work. Do not reset admission time when a garment moves to a repair bench or changes supervisor. Preserve the original time and add a separate repair event. A reviewer can then distinguish a short repair operation from a garment that waited a long time to receive it. The proposed record should support both the production decision and an honest explanation of customer delay.
Interlude · the cloth half-woven
Woven: sections 9.0 to 9.3Still to weave: 9.4 to 9.6
Read-out
Sourced from World Bank research and Enterprise Survey indicators: unconditional average revenue-based total factor productivity (TFP) gap for outage-exposed manufacturers across 84 developing countries, 2006 to 2019 (S36); share of surveyed Cameroon firms experiencing electrical outages in 2024 (S0301); and average value lost due to electrical outages as a share of sales for affected Cameroon firms in 2024 (S0302). Cross-country and sector-aggregate figures are context, not a single-factory garment forecast.
Unit: percent. Status: sourced. Sources: S36, S0301, S0302.
Read this exhibit in its chapter ↗9.4
50%
50%: Assumed mean queue wait 1 (illustrative mean queue wait in service-time units) · illustrative
MIT’s capacity lecture shows that the relationship between utilisation and waiting is strongly nonlinear under its queueing framework.
60%
60%: Assumed mean queue wait 1.5 (illustrative mean queue wait in service-time units) · illustrative
70%
70%: Assumed mean queue wait 2.3 (illustrative mean queue wait in service-time units) · illustrative
80%
80%: Assumed mean queue wait 4 (illustrative mean queue wait in service-time units) · illustrative
90%
90%: Assumed mean queue wait 9 (illustrative mean queue wait in service-time units) · illustrative
GFT illustrative M/M/1 model, not measured garment data. Independent exponential arrivals and service; one server; infinite buffer; stable utilisation below 1; mean service time = 1 arbitrary unit. Wq = rho/(1-rho). Assumed rho = 0.50, 0.60, 0.70, 0.80, 0.90 produces 1, 1.5, 2.3333, 4, 9. No recommended operating limit is inferred.
Unit: illustrative mean queue wait in service-time units. Status: illustrative. Sources: S23. Illustrative figures are teaching examples, not measured factory results.
9.4 · the bodyExhibit 42 above; its figures are illustrative where marked.
MIT's capacity lecture shows that the relationship between utilisation and waiting is strongly nonlinear under its queueing framework. The teaching distinguishes stable long-run queueing from short-run accumulation and notes the importance of arrival and service variability. It supports investigating congestion, not adopting a single utilisation target for every garment operation.[S23]
GFT's proposed interpretation is that unused capacity needs a purpose and a location. At a non-constraint, the team can use available time for authorised recovery, preparation, inspection or learning, subject to competence and safe methods. At the current constraint, lost productive time requires a different response because it can affect completed output. The same idle-time observation should therefore trigger a question about its cause and system consequence before it triggers an instruction to make more pieces.
The illustrative curve in Exhibit 42 uses an assumed single-server model with independent exponential arrivals and service, unlimited waiting space, no breakdowns, no changeovers and a mean service time of one arbitrary time unit. Its mean queue wait is utilisation divided by the remaining capacity fraction, multiplied by mean service time. These are constructed conditions chosen to show curvature; the curve is not calibrated to a garment line or to Cameroon. The lecture provides the queueing framework.[S23]
Use this illustrative curve as a prompt for observation rather than as a scheduling calculator. A proposed line review should distinguish work arriving irregularly from work taking irregular amounts of service. Record whether the queue forms before a difficult operation, after a material shortage or because an exit inspection is blocked. A common symptom, more waiting, can justify different actions. Increasing the admission allowance before identifying the cause could make the symptom larger without addressing it.
A protective queue should be explicit, counted and justified. GFT recommends placing it only where the trial shows it protects accepted output against a specific interruption. Its purpose is different from an unidentified pile left by habitual overproduction. Name the replenishment event, the response when its contents fall and the action when its upper limit is reached. When the interruption changes, reconsider the queue instead of preserving an inherited allowance indefinitely.
The team should also record occasions when admission is deliberately paused. A pause caused by a full permitted queue may indicate that the release rule is working. A pause caused by missing components may indicate a preparation failure. A pause that starves the constraint may indicate an unsuitable limit or slow replenishment. Separate these states before calculating an occupancy measure. Otherwise, a useful refusal can be reported as a loss and the supervisor may be encouraged to defeat the release rule.
GFT proposes reducing permitted WIP gradually during a controlled trial while keeping the observation method stable. Predetermine the conditions requiring a return to the previous allowance, including deteriorating accepted output, unresolved quality containment or unsafe handling. A reversal should retain the record of what happened. It is evidence about the operating conditions, not a reason to erase the trial. The objective is a defensible release design whose response is understood.
GFT recommends reviewing replenishment reliability before changing a protective allowance. Ask the upstream owner to demonstrate how a falling queue generates action, who supplies the next complete kit and what happens when that kit is unavailable. A queue can be correctly sized for a nominal replenishment process that nobody actually follows. The trial should therefore observe the response as well as the stock level, and retain failures to respond as evidence for redesign.
9.5
Read-out · sourced
Published 3 March 2022. Single Ethiopian shirt-factory case combining kanban, single-piece flow and related process-loss work. The abstract reports these reductions. Baseline counts, elapsed-time definitions and causal attribution are not supplied by these tiles. These are not GFT results or Cameroon targets.
Unit: reported reduction from the study baseline. Status: sourced. Sources: S31.
A published Ethiopian shirt-factory case combined kanban and single-piece flow with other process-loss interventions. Its abstract reports reductions in WIP of 48.5 percent and throughput time of 80.3 percent. It also describes work on bundle tickets and packing. These are reported results from the study setting, not GFT results, a Cameroon benchmark or an isolated estimate of the effect of kanban alone.[S31]
The sourced quantities appear in Exhibit 43. The case supports a practical hypothesis: a controlled change to release and transfer arrangements merits testing. Its reported reductions do not establish what a different factory should promise to a buyer or assume in a financing model. The authors themselves propose further work across different product types.[S31]
For a proposed Cameroon pilot, GFT recommends selecting a defined garment family with identifiable entry and exit events, then observing an unchanged release arrangement before altering it. Record style, size mix, crew, working-time basis, repairs, power interruptions and missing materials in both periods. Select the comparison window before reviewing the results. If the intended conditions cannot be matched, retain the observations but describe the exercise as a learning trial rather than a clean causal comparison.
The pilot should have a written intervention description. Specify whether it changes kit admission, permitted queue size, transfer bundle, handover inspection or more than one feature. If operational needs require simultaneous changes, record the package as such. Do not attribute the entire improvement to a card system when methods, staffing or equipment changed at the same time. A usable factory decision may concern the package, but its explanation must preserve that boundary.
A proposed measurement record should combine accepted output, time-weighted WIP, cohort residence, repairs, missed component matches and reported handling problems. Keep the oldest open work visible alongside averages. Report intervals affected by outages or style changes instead of silently excluding them. Compare the work awaiting entry and finished work awaiting dispatch as well, because an apparent improvement inside sewing could simply relocate the queue. The admission rule is only successful within the boundary actually demonstrated.
GFT recommends agreeing a decision before the trial begins: extend, revise or stop. Extension requires sustained good output and better flow without a deterioration in quality, worker conditions or delivery. Revision is appropriate when a specific interface or exception defeats the proposed rule. Stopping is appropriate when the design cannot operate safely or maintain required output under the tested conditions. These criteria make an unsuccessful experiment useful and limit pressure to report a favourable percentage.
Keep the records practical enough for the people who must use them after the trial team leaves. A supervisor should be able to reconstruct why an urgent kit was admitted, where it waited and what event released the next permission. The reviewer should be able to trace a claimed improvement back to the same definitions used for the baseline. That traceability is the proposed acceptance evidence; a photograph of an orderly line or a larger hourly tally is insufficient on its own.
For a proposed exception review, group deviations by the decision that produced them. An unauthorised admission, an incorrect kit, a missing exit and an unresolved repair each require a different owner. Close an exception only when its physical disposition and record agree. Repeated administrative closure without reconciliation should prevent extension of the trial, because the apparent inventory reduction would then depend on the quality of the bookkeeping rather than on demonstrated production flow.
GFT recommends preserving the trial's unsuccessful intervals in the handover record. A design that works only with its originator present has not yet shown that the ordinary shift team can maintain it. Ask the receiving supervisor to run the admission and reconciliation routine using the written instructions, while the trial lead observes rather than intervenes. Record ambiguities and revise the method before expanding the operating boundary.
9.6
A Lean Enterprise Institute practitioner account argues that spare capacity can be normal in a flow arrangement, and illustrates the failure of producing components when a stopped operation prevents saleable completion. This is an operational argument from practitioner experience, not a quantified causal trial. It is useful here because it challenges the assumption that every machine should keep producing regardless of what the line can finish.[S28]
The Institute's account of the Theory of Constraints also emphasises identifying the limiting factor and subordinating other activities to it before considering additional capacity. Its financial meaning of throughput concerns money generated through sales. In this chapter, physical throughput means garments per declared unit of time. GFT recommends retaining the units explicitly so that accepted line output is never mistaken for invoiced sales or cash received.[S30]
Consider an illustrative line whose upstream operation can prepare 8 kits per hour but whose constraint completes 5 garments per hour, with no losses and adequate downstream capability. Releasing all 8 creates a net accumulation of 3 garment equivalents per hour. Over an illustrative 4-hour interval, the accumulation is 12. The upstream operator may appear fully occupied while completed output remains limited to 5 per hour. This is assumed flow-balance arithmetic, not an observed factory result.
…whose upstream operation can prepare 8 kits per hour but whose constraint completes 5…
illustrative
If, in the same illustrative model, release instead matches completion at 5 per hour, upstream productive occupancy is 5 divided by 8, or 62.5 percent of its assumed capacity. That percentage is neither a target nor a judgement about the operator. It follows from the deliberately unequal capacities. The remaining time needs an authorised use consistent with skills and safety. Increasing release purely to raise occupancy would restore the inventory accumulation without increasing constraint capacity.
release instead matches completion at 5 per hour
illustrative
The arithmetic is reconciled in Exhibit 44. Unlike the sourced case tiles, these figures are illustrative assumptions. They isolate a simple mechanism and exclude outages, repair and mix changes. The comparison therefore establishes neither the necessary protective queue nor a labour requirement.
Read-out
| Use upstream capacity | 8 | 5 | 3 |
|---|---|---|---|
| Match accepted exits | 5 | 5 | 0 |
GFT assumed steady-rate comparison with upstream capacity 8 kits/hour and constraint capacity 5 garments/hour, adequate starting WIP and downstream capability, no losses or rework. Net WIP change = release minus completion. Over 4 hours the excess-admission case adds (8-5)*4 = 12 garment equivalents. Controlled release occupancy is 5/8 = 62.5 percent, not a target.
Unit: illustrative kits or garments per hour. Status: illustrative. Illustrative figures are teaching examples, not measured factory results.
GFT proposes reviewing incentives against this mechanism. A supervisor should not be penalised for honouring a full-queue refusal while another report rewards unauthorised pieces. Present accepted completion, admission, WIP change and unresolved exceptions together. Keep local output counts for diagnosis, but require the reviewer to explain their relationship to completed garments. The purpose is to make an overproduction instruction visible before it becomes a routine response to a low local efficiency score.
The proposed oee-line simulator should preserve that same logic. Let an increased release setting enlarge the queue when completion cannot rise; do not automatically translate higher local running time into additional accepted output. Identify all assumed capacities, quality boundaries and interruption conditions. Link the model to observed records only after reconciling their definitions. This is a recommendation for model behaviour, not a claim that a simulator or a factory implementation has passed those checks.
The handover to the next chapter is therefore concrete. Once a workable admission rule and transfer boundary have been demonstrated, test whether changeovers, preparation and layout prevent smaller transfers from being sustained. Keep their capacity effects visible rather than reopening unlimited release to conceal them. GFT's recommended operating decision remains the same: admit work that can progress, protect the actual constraint and judge the system by good completion and elapsed time, with its inventory and human consequences in view.
Check the record
Every knot is a source this chapter cites; select one to read its record, or filter the register by grade. Publication dates and inherited-source limits stay explicit. Read the paper's full evidence summary.
6 sources · 17 citation marks on this pageMost cited: S22, 5 times
Grade AGrade BGrade CKnot size: citations on this page
All 6 records shown
c09-S22 · Grade A
LAI Lean Academy Glossary
Little’s law strictly applies to long term averages of stable systems, i.e. ones which are not starting, stopping or surging.
2012. Register access date: 2026-10-30. Source provenance is supplied by the paper evidence register.
Open the recorded source ↗c09-S23 · Grade A
Class 3: Capacity Lecture
The relationship between waiting time and capacity utilization is strongly non-linear!
2004. Register access date: 2026-10-30. Source provenance is supplied by the paper evidence register.
Open the recorded source ↗c09-S24 · Grade B
Pull Production
Nothing is produced by the upstream supplier process until the downstream customer process signals a need.
2020-07-02. Register access date: 2026-10-30. Source provenance is supplied by the paper evidence register.
Open the recorded source ↗c09-S28 · Grade B
Ask Art: How One-Piece-Flow Supports Quality
Having excess capacity is in fact normal and actually a good thing.
2016-01-13. Register access date: 2026-10-30. Source provenance is supplied by the paper evidence register.
Open the recorded source ↗c09-S30 · Grade B
What is the Theory of Constraints, and How Does it Compare to Lean Thinking?
TOC advocates beginning with the constraint that most limits Throughput.
2006-09-08. Register access date: 2026-10-30. Source provenance is supplied by the paper evidence register.
Open the recorded source ↗c09-S31 · Grade A
Process-loss Optimization through the Implementation of Kanban & Single-piece Flow: A Case Study in an Ethiopian Garment Factory
The WIP and throughput time were reduced by 48.5% and 80.3% respectively.
2022-03-03. Register access date: 2026-10-30. Source provenance is supplied by the paper evidence register.
Open the recorded source ↗S24 Lean Enterprise Institute. 2020-07-02.
S22 Massachusetts Institute of Technology, MIT OpenCourseWare. 2012.
S23 Massachusetts Institute of Technology, MIT OpenCourseWare. 2004.
S31 Journal of Textile and Apparel, Technology and Management, North Carolina State University. 2022-03-03.
S28 Lean Enterprise Institute. 2016-01-13.
S30 Lean Enterprise Institute. 2006-09-08.
Keep the thread · chapter 09 of 14
Move the shuttle along the rack to preview any chapter: its sections, exhibits and reading time. Draft web edition. This research is not investment, legal or tax advice; illustrative models are not measured factory results.
The paper's rack · 14 boltsBolt length: reading time
01Part I: Measure the factory that actually exists
A garment factory creates value when the whole line converts available time and skill into accepted output. Installed machines and busy stations are evidence of resources, not proof of delivery.
02Part I: Measure the factory that actually exists
A single OEE score can start a conversation, but only its components and loss codes can tell a supervisor what to fix.
03Part I: Measure the factory that actually exists
A factory cannot interpret downtime until it has separated legal hours, planned production, external interruptions and losses inside the line.
04Part I: Measure the factory that actually exists
SMV translates garment work content into earned minutes, but it must sit beside OEE and quality rather than replace them.
05Part II: Balance work to the constraint
Balancing begins with operation precedence and required output, then tests how work can be shared without violating machine, skill or quality constraints.
06Part II: Balance work to the constraint
A workstation assignment is feasible only when the operator can perform the method safely and repeatedly at the assumed cycle and quality.
07Part II: Balance work to the constraint
A balance that ignores machine availability, skill or precedence can look efficient on screen and fail on the floor.
08Part II: Balance work to the constraint
A balance calculated for one style and crew decays as absence, learning, defects and order mix change the work.
09Part III: Design flow instead of local busyness
Balanced work needs controlled release. A line shortens delivery time when it admits work that downstream operations can use, makes queues visible and protects good output at its constraint.
10Part III: Design flow instead of local busyness
Small batches become practical when style changes are short, methods are standard and the layout lets qualified operators share work around the constraint.
11Part III: Design flow instead of local busyness
Defects consume capacity twice, first through lost good output and then through rework, queues and repeated interruption.
12Part IV: Turn measurement into sustained performance
A technically balanced line still fails when learned operators are absent, hours are unsustainable or electricity interrupts production.
13Part IV: Turn measurement into sustained performance
The financial benefit is not a higher dashboard score; it is more good garments at acceptable labour, energy, capital and working-capital cost.
14Part IV: Turn measurement into sustained performance
Release the next commitment only when a measured trial shows that good throughput, safe work and controlled flow survive ordinary operating conditions.