Introduction
This coursework applies the Define, Measure, Analyse, Improve and Control (DMAIC) framework to a lean process improvement problem at Kingsvale Hydraulics, a hypothetical tier-two components manufacturer in Dandenong South, Victoria. The plant machines and assembles cast aluminium hydraulic manifolds for agricultural and mining equipment original equipment manufacturers (OEMs) and employs 58 people across two shifts. The brief requires a quantified current-state analysis, a diagnosis of waste, a costed improvement plan and a control strategy. The analysis is organised by DMAIC phase: takt and cycle times, overall equipment effectiveness (OEE) at the constraint, lead time via Little’s Law, waste against Ohno’s (1988) seven categories, and a sequenced plan with owners and expected gains.
Define
Operating context
Australian manufacturers such as Kingsvale operate under sustained margin pressure. Manufacturing remains one of the country’s larger employing industries, with the Australian Bureau of Statistics (2024) recording approximately 900,000 people employed across the sector, yet the Australian Industry Group (2024) has reported readings below the 50 point expansion threshold on its Australian Performance of Manufacturing Index for much of the recent period, reflecting weak orders alongside elevated energy and labour costs. The Productivity Commission (2023) locates much of the remaining opportunity for productivity growth in firm-level process capability rather than capital deepening. For a plant competing on delivery reliability, internal process improvement is therefore the only realistic lever on margin.
Problem statement and scope
Kingsvale’s principal OEM customer has moved to a weekly call-off schedule and measures suppliers on delivery-in-full-on-time performance. Over the last two quarters the plant has achieved 91% against a contractual target of 98%, has absorbed recurring Saturday overtime, and holds more than five weeks of inventory. The project is scoped to the manifold machining and assembly line, from billet receipt to finished goods despatch. The objective is to lift good output to the contracted 300 units per day without overtime, and to cut production lead time by at least 50% within twenty-six weeks.
Measure
Takt time and line balance
Takt time is the rate at which the line must produce to match demand, and is the reference against which every station is assessed (Rother & Shook 2003). The plant runs two eight-hour shifts across five days. Each 480 minute shift loses 20 minutes of rostered breaks and a 10 minute start-up and 5S routine, giving 450 minutes of available production time per shift and 900 minutes per day. Contracted demand is 300 units per day.
Takt time = available production time / customer demand = 900 minutes / 300 units = 3.0 minutes per unit (180 seconds).
Table 1 records cycle time, changeover, first-pass yield and staffing for the six stations.
Table 1: Current-state process data for the manifold line (mean of 30 timed cycles per station).
| Station | Operation | Cycle time (min) | Changeover (min) | First-pass yield | Operators |
|---|---|---|---|---|---|
| 10 | Billet saw and prep | 1.5 | 12 | 99.4% | 1 |
| 20 | CNC mill, Op 10 | 1.8 | 25 | 98.9% | 1 |
| 30 | CNC mill, Op 20 | 2.2 | 45 | 95.0% | 1 |
| 40 | Deburr and wash | 1.6 | 5 | 99.6% | 1 |
| 50 | Pressure and leak test | 1.3 | 8 | 98.2% | 1 |
| 60 | Assembly and pack | 1.9 | 10 | 99.5% | 1 |
| Line total | 10.3 | 105 | 90.9% | 6 | |
Two measures follow from Table 1. Rolled throughput yield, the probability that a unit passes every station without rework, is the product of the individual yields: 0.994 x 0.989 x 0.950 x 0.996 x 0.982 x 0.995 = 0.909, or 90.9%. Line balance efficiency = total cycle time / (stations x bottleneck cycle time) = 10.3 / (6 x 2.2) = 10.3 / 13.2 = 78.0%. Every station cycles below the 3.0 minute takt, which appears to indicate adequate capacity, yet Station 30 is the constraint: it is the slowest station and carries both the longest changeover and the lowest yield.
Overall equipment effectiveness at the constraint
Because throughput is governed by the constraint, capacity analysis is concentrated there (Goldratt & Cox 2016). OEE decomposes the gap between planned and actual output into availability, performance and quality losses (Nakajima 1988). Planned production time at Station 30 is 900 minutes per day, and downtime comprises two changeovers of 45 minutes plus 90 minutes of unplanned stoppages for tool breakage, coolant faults and awaiting inspection, a loss of 180 minutes. In the resulting 720 minutes of operating time the machine produced 280 units, of which 266 passed first time.
- Availability = operating time / planned production time = (900 – 180) / 900 = 720 / 900 = 0.800, or 80.0%.
- Performance = (ideal cycle time x total count) / operating time = (2.2 x 280) / 720 = 616 / 720 = 0.856, or 85.6%.
- Quality = good count / total count = 266 / 280 = 0.950, or 95.0%.
- OEE = availability x performance x quality = 0.800 x 0.856 x 0.950 = 0.650, or 65.0%.
An OEE of 65.0% sits well below the 85% benchmark Nakajima (1988) proposed for a well-run discrete process, although Muchiri and Pintelon (2008) caution that the composite is useful only when its three components are read separately, as above. The consequence is an inflated effective cycle time: effective cycle time = ideal cycle time / OEE = 2.2 / 0.650 = 3.38 minutes per unit, 0.38 minutes above takt. Good output is therefore 900 x 0.650 / 2.2 = 266 units per day against a requirement of 300, a shortfall of 34 units per day or 170 per week. Recovering it consumes 170 x 3.38 = 575 minutes, approximately 9.6 hours of Saturday overtime. Under the Manufacturing and Associated Industries and Occupations Award, Saturday overtime is paid at 150% for the first three hours and 200% thereafter (Fair Work Ombudsman 2024), so at A$34.00 per hour for a crew of seven the weekly cost is 7 x [(3 x 51.00) + (6.6 x 68.00)] = A$4,212.60, or approximately A$194,000 across 46 operating weeks.
Value stream and production lead time
Figure 1 presents the current-state value stream, with inventory at each interface and the constraint marked.
Work-in-process across the seven inventory points in Figure 1 totals 3,600 + 320 + 1,150 + 240 + 180 + 210 + 1,500 = 7,200 pieces. Little’s Law relates inventory, throughput and lead time in any stable system (Hopp & Spearman 2011): lead time = work-in-process / throughput = 7,200 / 266 = 27.1 production days. Value-adding work is a negligible fraction of that interval: process cycle efficiency = 10.3 / (27.1 x 900) = 10.3 / 24,360 = 0.04%. Ratios of this order are characteristic of batch-and-queue manufacturing and locate the opportunity in waiting rather than in working faster (George 2002).
Analyse
The data were interrogated against Ohno’s (1988) seven wastes, extended to include the underutilisation of employee capability that Liker (2021) treats as the eighth. Table 2 records each category, the observation at Kingsvale and its quantified impact.
Table 2: Waste (muda) identification for the manifold line, with quantified impact.
| Waste category | Observation at the plant | Quantified impact |
|---|---|---|
| Overproduction | Station 20 runs batches of 250 to amortise a 25 minute changeover | 1,150 pieces queued at the constraint, 3.8 days of cover |
| Waiting | Leak test cycles in 1.3 min against a 3.0 min takt | 1.7 min idle per unit, about 452 min per day |
| Transport | Pallet-load forklift move of 38 metres to the wash cabinet | 28 movements per day across a pedestrian aisle |
| Over-processing | 100% manual gauging of a bore at Cpk 1.67 | About 84 min of constraint capacity per day |
| Inventory | Seven inventory points totalling 7,200 pieces | 27.1 day lead time; about A$396,000 of working capital |
| Motion | Constraint operator walks 11 metres per cycle to crib and gauge | About 3.1 km per day, a fatigue and time loss |
| Defects | Bore oversize and surface finish rejects at Mill Op 20 | 14 units per day, 5.0% of output, about A$224,000 a year |
| Underutilised talent | No problem-solving routine; 4 of 12 operators maintenance-trained | Improvement activity dependent on two engineers |
Read together, the three largest items share a root cause. The 45 minute changeover at Station 30 makes small batches uneconomic, which drives overproduction at Station 20, which builds the 1,150 piece queue that is the largest single component of the 27.1 day lead time; the same changeover accounts for half of all constraint downtime. A five-why exercise on the bore oversize defect traced it to tool wear compensation applied by operator judgement rather than by measured offset, itself a consequence of gauging performed remote from the machine. The waste categories are therefore symptoms of setup time, absent standardised work, and inventory used as a buffer against variability.
Improve
Table 3 sets out the intervention plan. Sequencing follows the theory of constraints: exploit the constraint through setup reduction and maintenance, subordinate the line to it through pull, and only then elevate capacity with capital (Goldratt & Cox 2016).
Table 3: Improvement plan with expected gains, owners and scheduling.
| Intervention | Expected gain | Owner | Weeks |
|---|---|---|---|
| SMED at Station 30 with pre-set tooling and external setup | Changeover 45 to 15 min; 60 min per day returned | Manufacturing engineer | 1-8 |
| Autonomous maintenance and tool-life management at Station 30 | Unplanned stoppages 90 to 50 min; availability 80.0% to 87.8% | Production supervisor | 4-16 |
| In-machine probing plus automated post-process gauge at Station 40 | Performance 85.6% to 89.1%; 84 min recovered | Quality manager | 6-14 |
| A3 problem solving on the two dominant scrap modes | Quality 95.0% to 96.9%; scrap 14 to 10 units per day | Quality manager | 6-26 |
| Standardised work and operator certification at all stations | Cycle time variation reduced; full training coverage | Line team leader | 8-18 |
| Kanban pull between stations with a 300 piece constraint buffer | Internal WIP 2,100 to 660 pieces; overproduction eliminated | Materials manager | 10-24 |
| U-shaped cell re-layout removing the forklift transfer to wash | Transport eliminated; operator walk 11 to 4 metres | Manufacturing engineer | 12-22 |
| Weekly call-off kanban with the Victorian billet supplier | Billet stock 3,600 to 1,200; finished goods 1,500 to 600 pieces | Procurement officer | 14-26 |
The capacity effect of the first four interventions is modelled directly. Downtime falls from 180 to 110 minutes per day, so availability = (900 – 110) / 900 = 790 / 900 = 87.8%. With gauging removed from the manned cycle, total count rises to 320 units in 790 minutes of operating time, so performance = (2.2 x 320) / 790 = 704 / 790 = 89.1%, and defect reduction lifts quality to 310 / 320 = 96.9%. OEE = 0.878 x 0.891 x 0.969 = 0.758, or 75.8%, a gain of 10.8 percentage points. Good output becomes 900 x 0.758 / 2.2 = 310 units per day, 3.3% above requirement, and effective cycle time falls to 2.2 / 0.758 = 2.90 minutes, inside takt. Saturday overtime is no longer structurally required.
The pull interventions deliver the lead-time gain. Future-state work-in-process is 1,200 billets, 120 + 300 + 90 + 60 + 90 pieces between stations and 600 finished units, a total of 2,460 pieces, so future lead time = 2,460 / 300 = 8.2 days. The reduction is 27.1 – 8.2 = 18.9 days, or 69.7%, ahead of the 50% objective, and process cycle efficiency improves to 10.3 / (8.2 x 900) = 0.14%. The plan avoids approximately A$194,000 of annual overtime, A$64,000 of scrap at a burdened cost of A$64 per unit and about A$52,000 of carrying cost at a 20% holding rate, a combined benefit near A$310,000 a year plus a one-off working capital release of roughly A$260,000. Against an outlay of A$180,000, simple payback = 180,000 / 310,000 = 0.58 years, approximately seven months.
Control
Netland (2016) observes that lean programs fail through erosion rather than through the initial improvement, so four control mechanisms are proposed. Standardised work sheets are posted at the point of use and owned by the line team leader, since a standard held by engineering and not by operators is not a control (Liker 2021). OEE at Station 30 is captured from the machine controller and reviewed at a daily fifteen minute meeting at the line, with availability, performance and quality shown separately so that a fall in any component is visible the same day. Control charts on bore diameter, with a reaction plan naming who acts and at what trigger, replace the judgement-based tool offsets. The kanban loops are audited fortnightly, because inventory creep is the earliest reliable signal that pull discipline is failing.
Sustainment is also anchored in existing governance. The revised standards, competency records and reaction plans are incorporated into the plant’s documented information under AS/NZS ISO 9001, so that they fall under internal audit and management review rather than depending on the improvement team (Standards Australia 2016). The re-layout also removes a forklift route crossing a pedestrian aisle. A monthly review of the categories in Table 2 keeps the diagnostic live, so that when the constraint migrates to Station 60, as it will once Station 30 runs at 2.90 minutes, the next cycle begins from measured evidence.
Conclusion
Applying DMAIC to the Kingsvale manifold line shows that its delivery failure is not a capacity problem in the conventional sense. Every station cycles inside the 3.0 minute takt, yet an OEE of 65.0% at the constraint inflates the effective cycle time to 3.38 minutes and creates a structural shortfall of 34 units per day that overtime has concealed at a cost approaching A$194,000 a year. The waste analysis traced the dominant losses to a single 45 minute changeover, which drives batch sizing, the constraint queue and half of all downtime. The plan lifts OEE to 75.8% and good output to 310 units per day, cuts lead time from 27.1 to 8.2 days, and returns approximately A$310,000 a year on an outlay of A$180,000. The wider point for Australian manufacturers facing the cost pressures reported by the Australian Industry Group (2024) is that these gains required no additional machine tools. The principal risk is reversion, which is why the controls are embedded in daily routine and in the certified quality system rather than in the project itself.
References
Australian Bureau of Statistics 2024, Australian industry, 2022-23, ABS, Canberra.
Australian Industry Group 2024, Australian Industry Group Australian Performance of Manufacturing Index, Ai Group, Sydney.
Fair Work Ombudsman 2024, Pay guide: Manufacturing and Associated Industries and Occupations Award, Australian Government, Canberra.
George, ML 2002, Lean Six Sigma: combining Six Sigma quality with lean speed, McGraw-Hill, New York.
Goldratt, EM & Cox, J 2016, The goal: a process of ongoing improvement, 30th anniversary edn, North River Press, Great Barrington.
Hopp, WJ & Spearman, ML 2011, Factory physics, 3rd edn, Waveland Press, Long Grove.
Liker, JK 2021, The Toyota Way: 14 management principles from the world’s greatest manufacturer, 2nd edn, McGraw-Hill, New York.
Muchiri, P & Pintelon, L 2008, ‘Performance measurement using overall equipment effectiveness (OEE): literature review and practical application discussion’, International Journal of Production Research, vol. 46, no. 13, pp. 3517-3535.
Nakajima, S 1988, Introduction to TPM: total productive maintenance, Productivity Press, Cambridge.
Netland, TH 2016, ‘Critical success factors for implementing lean production: the effect of contingencies’, International Journal of Production Research, vol. 54, no. 8, pp. 2433-2448.
Ohno, T 1988, Toyota production system: beyond large-scale production, Productivity Press, Cambridge.
Productivity Commission 2023, 5-year productivity inquiry: advancing prosperity, Productivity Commission, Canberra.
Rother, M & Shook, J 2003, Learning to see: value stream mapping to add value and eliminate muda, Lean Enterprise Institute, Cambridge.
Standards Australia 2016, AS/NZS ISO 9001:2016 Quality management systems: requirements, Standards Australia, Sydney.