Current-State Map With a Hidden Constraint and a Tempting Inventory Pile
An order-fulfilment plant builds one product family: 600 units/day on one 450-minute shift (27,000 seconds available). Stamping has a 40-second cycle, 90% uptime and 25-minute changeovers, with 6,000 units of raw material on hand. Welding follows with a 55-second cycle, 96% uptime and 10-minute changeovers, with 3,600 units between stamping and welding. Final assembly builds three variants at a 35-second cycle, 99% uptime, with 1,200 units between welding and assembly. Packing follows, with 2,400 units of finished goods. Material is pushed on weekly ERP schedules and each department is measured on its own efficiency. The plant has completed 38 improvement projects this year - every one delivered its local target, none was measured against lead time, and none crossed a department boundary. The value stream manager wants to start with the 6,000 units of raw material because it is the biggest pile. Welding runs Saturday overtime most weeks to catch up.
1. Calculate takt time, the uptime-adjusted cycle at each process, total lead time and process cycle efficiency. Identify the constraint and state what that implies about where improvement effort belongs.
Read the worked answer
Takt = 27,000 / 600 = 45 s.
Uptime-adjusted cycle = cycle time ÷ uptime:
- Stamping: 40 / 0.90 = 44.4 s (just under takt - no margin)
- Welding: 55 / 0.96 = 57.3 s - above takt, so welding alone cannot serve demand in 450 minutes: 57.3 × 600 = 34,380 s = 573 min, i.e. 123 minutes of overtime per day, which is exactly the Saturday overtime observed.
- Assembly: 35 / 0.99 = 35.4 s, comfortably under takt.
Lead time = total inventory ÷ daily demand = (6,000 + 3,600 + 1,200 + 2,400) / 600 = 13,200 / 600 = 22 days = 594,000 s. Process time = 40 + 55 + 35 = 130 s (plus packing). PCE = 130 / 594,000 ≈ 0.022%.
Welding is the constraint, and every hour of welding loss costs about 3,600 / 57.3 ≈ 63 units of output - more than any other hour in the plant. Improvement effort therefore belongs at welding first (uptime, changeover, minor stops), and on the scheduling rules that starve or choke it. Nothing in the data supports effort at 38 unfocused local projects.
2. Evaluate the value stream manager's plan to attack the raw material inventory first. What sequence would you recommend and why?
Read the worked answer
The plan mistakes the biggest pile for the biggest lever. Raw material is 10 days, but it is two processes upstream of the constraint, and reducing it requires supplier and purchasing changes rather than process change. Worse, cutting it first is actively risky: stamping has only 44.4 s effective cycle against a 45 s takt, so any stamping interruption propagates straight to welding, and welding is the constraint. You would trade cheap inventory for expensive constraint downtime.
Recommended sequence:
- At the constraint: raise welding uptime and cut the 10-minute changeover further, and protect it with a small, controlled buffer of stamped parts sized to cover stamping's demonstrated variability - not a fixed 6 days.
- Immediately downstream of the constraint: the 2 days between welding and assembly is the queue the plant most controls and the one that hides welding's stoppages. Shrink it once welding is stable.
- Then work upstream: FIFO-cap the stamping-to-welding lane, then attack changeover so batch size falls, then renegotiate raw material in smaller, more frequent deliveries with the supplier.
- Measurement: every project from now on reports lead time, WIP days and OTIF for the stream, not just its local target - the absence of that metric is why 38 successful projects changed nothing.
General rule: sequence by constraint and by controllability, not by pile size, and never remove a buffer that protects the constraint before the feeding process is stable.
3. Design the future-state flow between welding and assembly. Would you use a supermarket pull or a FIFO lane, and what data justifies the choice?
Read the worked answer
Choose a supermarket of welded subassemblies, held by variant, with pull from assembly - and cap it.
Reasoning from the data: welding is the constraint (57.3 s effective vs 45 s takt) and produces three variants for assembly. A size-changeover process feeding a three-variant downstream is exactly the case where a supermarket pays: it lets the constraint run in sensible changeover lots, lets assembly pull the variant the customer ordered, and decouples assembly from welding's 4% downtime. Size it from the data - cover welding's demonstrated downtime plus one replenishment interval, say about one shift of assembly demand (~200 units total, split by variant) - not from a rule of thumb. FIFO is credible only where changeover is near zero and mix is stable, which is not true here.
Caveats: a supermarket downstream of the constraint can insulate welding from the true demand signal, so the supermarket must be capped, replenished strictly by pull, and welding scheduled as the pacemaker with a levelled variant sequence. Watch the supermarket level daily; if it trends up, the schedule is batching or demand has dropped.
Between stamping and welding, use a capped FIFO lane: one variant, coupled processes with near-identical effective cycles, and short distance. A supermarket there would add handling for variety that does not exist.