Takt Time vs a Capital Request in a Batch-Scheduled Press Shop
A 210-person precision stamping plant runs two 450-minute shifts, five days a week. Customer demand is steady at 640 assemblies per day, and the customer has moved to daily truck departures with a 24-hour order freeze. The plant still schedules in two-week batches and changes over the 800-ton press four times a month at roughly four hours each. The bottleneck press cycles at 78 seconds with measured OEE of 62%, driven by die crashes and short stops. Operators receive a monthly bonus tied to machine utilization above 85% and to individual efficiency, so they run ahead and stack WIP between presses; there are nine days of finished goods in the shipping lane despite the daily pull. Last quarter the plant missed two delivery windows and paid premium freight, and the quality manager found a five-day-old defect that had escaped to the customer. The operations director's proposal is to buy a second press.
1. Using takt time, show whether the plant can meet demand with its current assets, and state what the operations director should do instead of buying a press.
Read the worked answer
Available time = 2 × 450 min = 900 min = 54,000 s. Takt = 54,000 / 640 = 84.4 s.
The press nominally cycles at 78 s, which is under takt, so on paper one press is enough - but only if it runs essentially all available time. At the measured 62% OEE it delivers 54,000 × 0.62 / 78 ≈ 429 good units/day, a shortfall of 211. Even a strong 85% OEE gives ≈ 588 units, still short. Break-even requires 640 × 78 / 54,000 = 92.4% OEE with no overtime.
So this is not yet a capacity decision; it is a decision about whether the losses are recoverable. A second press adds roughly 690 units/day of nominal capacity the plant does not need, plus $1.5–2M and floor space, and it will be under-loaded as soon as the constraint moves elsewhere.
The defensible sequence: (1) a 90-day focused attack on the six big losses at the press - die crashes, short stops, changeover - with a stop-the-line and downtime-reason discipline; (2) SMED on the four-hour changeover, which both buys capacity and makes smaller batches feasible; (3) overtime as an explicit interim bridge; (4) re-open capex at the 92.4% milestone. Note the required OEE leaves no buffer for demand growth, so the press question returns in two to three years - with data, not now.
2. Identify the waste created by the utilization and individual-efficiency bonus, trace its consequences through the value stream, and propose what the measure should be replaced with.
Read the worked answer
The bonus pays for overproduction, which is the waste that generates the others. Running ahead fills the presses with work nobody ordered yet, and because the bonus is per machine and per person, each operator optimizes his own station against a boundary he does not control.
Consequences trace in a chain:
- WIP between presses grows, so lead time grows, so the customer order freeze is exceeded - hence the daily pull cannot actually pull.
- Defects are detected days later (the five-day-old escape), because parts sit in queues before inspection; the defect is then scattered across a large batch, so containment and sorting cost multiply.
- The variability that actually stops the plant (die crashes) is hidden, because the extra inventory keeps downstream busy while the press is down. The push system masks the constraint.
- Changeover is punished: every setup minute costs the operator bonus, so four-hour changeovers survive and batch size stays large.
Replacement measures: put OEE-with-quality at the constraint only; on-time-in-full and lead time for the stream; first-pass yield; changeover time and batch size; and a plant-level gain-share rather than individual efficiency, so no one is paid to build inventory. Individual efficiency at a shared resource is the specific measure to retire first.
3. The plant manager proposes running all five presses at 100% utilization and building to forecast to 'absorb variability'. Evaluate that proposal.
Read the worked answer
It is the classic local-optimum trap. Throughput of a line is set by the constraint, not by the sum of local utilizations. Only the bottleneck press converts extra hours into extra shipped units; at any non-bottleneck press, 100% utilization converts hours into inventory, which is cash and lead time, not throughput.
Building to forecast makes it worse in four ways. It decouples production from the actual daily pull the customer has already imposed, so error in the forecast becomes inventory or shortage. It lengthens lead time, so the plant responds later to mix changes. It hides the die-crash losses that are the real cause of missed windows. And it consumes the very capacity that SMED and reliability work would otherwise free for genuine demand growth.
A better framing: protect the constraint, starve nothing, and let non-bottlenecks idle. Set the constraint's schedule from takt and the 24-hour freeze, run controlled buffers only where they protect the constraint and the customer, and measure flow (lead time, OTIF, WIP days) rather than utilization. If the plant wants a utilization target, the honest one is 'the constraint runs at its demonstrated best; everything else runs to pull'. That statement is testable, whereas 'everyone at 100%' cannot be satisfied without overproduction.