Throughput Gap on a Five-Station Line
A five-station manual assembly line runs one 480-minute shift. Daily demand is 120 units. Measured station cycle times are 3.2, 4.4, 3.0, 4.1 and 2.9 minutes. Rework runs at 12% of output and returns through the same stations; the dominant defect is a connector seating error detected at station 4 but originating in work performed at station 2. The line runs 118 units on a good day and 96 to 104 on most days, with 60 to 80 units of work-in-process on the floor. The operations lead wants to add a sixth technician at station 2.
1. Do the capacity arithmetic. Is a sixth technician at station 2 the right intervention?
Read the worked answer
Takt time is 480 / 120 = 4.0 minutes. Station 2 at 4.4 minutes is the constraint, giving 480 / 4.4 ≈ 109 units per day gross - already below demand before rework is counted.
Rework changes the picture substantially. To ship 120 good units you must start about 136, and every reworked unit re-occupies each station. Loading the constraint for rework gives an effective 4.4 × 1.12 ≈ 4.93 minutes, or roughly 97 units per day. That matches the observed 96 to 104 much better than the gross figure does, which is a good sign the model is right.
Adding a technician at station 2 does not help while station 4 (4.1 min) is the next constraint. It will increase output at station 2 into a queue that station 4 cannot clear, raise WIP above its current 60 to 80 units, and lengthen the feedback loop. The intervention is misdirected because it is aimed at the station that looks slow rather than the system constraint.
The correct move first is rebalancing toward a 3.7-minute target: about 0.7 minutes of station 2 work (a subassembly that does not need to be done in line) moves to stations 1 and 3, and about 0.4 minutes of station 4 work moves to station 5. The stations then sit near 3.4, 3.7, 3.5, 3.7 and 3.3. Capacity becomes 480 / 3.7 ≈ 130 gross, about 116 net of 12% rework - closer but still short of 120, which tells you rebalancing alone does not close it. Defect reduction must be part of the answer: at 6% rework the same line yields roughly 123 net.
2. Explain the interaction between the 12% rework and the 60 to 80 units of WIP, and what it implies about control.
Read the worked answer
The defect originates at station 2 and is detected at station 4 - two stations and roughly two hours of queue time later. So every escape consumes capacity at stations 3 and 4, and the rework loop competes with new production for the same constrained minutes. In effect 12% of the line's capacity is being spent on work that will be done twice.
WIP of 60 to 80 units against a 4-minute takt is roughly two hours of inventory on the floor. That is the mechanism that hides the constraint: station 2 can keep producing into a buffer, so station 4's starvation or saturation is invisible until the buffer drains or floods, and nobody can see the actual bottleneck from the floor. It also lengthens the defect feedback loop to the point where station 2 cannot connect its own work to the connector failures found downstream.
The control implications are concrete. Put detection at the source: a verification or error-proofing step at the station 2 to 3 interface so the connector condition is caught before two more stations of value are added to a unit. Cap WIP between stations at around 10 units and stop the line on defect rather than building ahead. Both changes will initially reduce daily output and look like a step backwards on the metric; that is expected, and it should be agreed in advance so the drop is not read as failure.
3. What do you do in the first two weeks, and in what order?
Read the worked answer
Week one is measurement and stability, not intervention. Verify the station times independently, because the numbers currently in use may be nominal rather than observed, and a rebalancing decision built on wrong times simply relocates the problem. Establish a defect Pareto by station of origin rather than station of detection - the connector issue needs to be confirmed as the top mode, not assumed. Institute the WIP caps and the defect-stop rule, and separate rework from new production in the metrics so that rework hours stop being invisible inside the daily count.
Week two is root cause and the first countermeasure. Run the connector seating problem down to a physical cause - fixture, torque, technique, part variation - before writing any training action, because training is not a countermeasure if the cause is a fixture. Re-measure the line with the WIP caps in place; the constraint often moves once queues are controlled.
Only then rebalance, and only then consider headcount. If a sixth person is added, place them on the constraint after rebalancing has moved it - as a floater covering the current bottleneck and the rework loop - not on the station that happened to look slowest on day one. Sequence the reasons explicitly so that when output dips during stabilisation, the plan already explains why.