Two Ways to Fix Line 3
Kestrel Machining runs a six-machine CNC cell producing one valve-body family. Weekly demand is 1,500 pieces. Run time is 2.8 minutes per piece, and the cell has 120 machine-hours available per week across two shifts. Lot size is 300 pieces, which produces five changeovers per week; each changeover takes 52 minutes. Every-planning-interval, the time to cycle through the full product mix, is currently about two weeks, and the customer has begun asking for weekly ordering. Two proposals are on the table. The manufacturing engineer wants a seventh lathe at $310,000, claiming a two-year payback on added capacity. A setup team has proposed a three-month SMED effort budgeted at $15,000. A time study of the changeover breaks down as follows: 22 minutes searching for tooling and fixtures in a shared crib, 11 minutes mounting and removing dies, 8 minutes trial adjustment to get the first part in tolerance, 6 minutes with the machine stopped for paperwork and unloading the previous lot, and 5 minutes of actual cleanup. The plant manager asks which proposal to fund.
1. Evaluate the two proposals using the available capacity numbers. Is capacity the constraint?
Read the worked answer
Do the arithmetic before discussing either proposal.
Load: 1,500 pieces × 2.8 min = 4,200 min = 70.0 hours of run time. Changeovers: 5 × 52 min = 260 min = 4.3 hours. Total load = 74.3 of 120 available hours, about 62%. Even allowing for breaks, PM, and minor stops, roughly 40 hours per week are unaccounted for.
Conclusion: the cell is not capacity-constrained, so the $310,000 lathe cannot pay back on capacity. Worse, adding a machine changes nothing about lot size: at 300 pieces and 52-minute changeovers, the planning interval stays near two weeks, work-in-process stays high, and lead time is unchanged. The engineer is answering a lead-time complaint with a capacity purchase.
On the changeover data: 22 of the 52 minutes are not changeover work at all - they are waiting for tools because the crib is shared. That is external setup discipline and kanban at the tool crib, not capital. Another 6 minutes are paperwork and unloading that could be done during the run.
Funding recommendation: fund the $15,000 SMED effort and a modest tool-crib reorganization, and hold the lathe decision until the changeover data is real. Set the revisit condition in advance so the team knows the decision is not permanent: revisit capital only if, after two events, changeover is at or below 25 minutes, lot size has dropped to 150, and the cell still exceeds 85% load.
2. Describe the sequence of SMED steps you would run and what you would expect the changeover standard to contain.
Read the worked answer
Run the classic sequence, in order, without skipping ahead to hardware.
1. Observe and record. Film a changeover on second shift, with the team watching the film, and separate every minute into internal (machine stopped) and external (can be done while running). The 6 minutes of paperwork and lot unloading move to external immediately. So do pre-staging of the next lot's fixtures and gauges.
2. Convert internal to external. The 22 minutes of crib search disappear if the next lot's tooling is kitted on a cart before the machine stops. Give the crib a min/max and a two-bin kanban so a setup technician never walks into a shortage.
3. Streamline what remains. Quick-release clamps instead of bolts, standardized die heights, torque markers, a documented sequence with a numbered setup sheet.
4. Eliminate adjustment. The 8 minutes of trial adjustment exist because the machine has to be dialed in by feel. A setting master, a first-piece gauge at the machine, and a repeatable zero point take this toward one or two minutes.
The resulting standard should contain: the numbered step list with target minutes per step, the cart contents by product family, the external checklist that must be complete before the machine stops, the first-piece verification method, and the recorded actual time for every changeover, visible on the cell board. The operators and the setup technician should author the sheet, not the engineer. A standard they did not write is a standard they will not follow when the shift gets tight.
3. The first kaizen event takes changeover from 52 to 38 minutes. The sponsor wants to cancel the second event and buy the lathe because the first event "only got 27%." What is your response?
Read the worked answer
First, reframe the result. 52 to 38 minutes is 14 minutes recovered per changeover - 70 minutes per week across five changeovers, roughly 58 hours a year from one three-day event, at almost no cost. That is a normal first-pass SMED result, and the event deliberately tackled the easy external work first. The 22 minutes of crib search and the 8 minutes of trial adjustment are still untouched, and those are the second event's target.
Second, test the proposal against the real constraint. At 38 minutes and a lot size of 150, the cell runs 10 changeovers per week: 380 minutes, or 6.3 hours, on top of 70 hours of run time - 76.3 of 120 hours. Planning interval drops to about one week, which is exactly what the customer asked for. The cell still has over 40 hours of headroom. Nothing in that arithmetic supports a $310,000 machine.
Third, protect the decision from being made on disappointment rather than data. Write the decision rule down and get the sponsor to agree to it: complete event two; if changeover reaches 25 minutes or less and lot size is 150, the lathe stays unfunded. If the cell still exceeds 85% load at that point, reopen capital with the measured numbers rather than the engineer's estimate.
Finally, ask why a 27% gain reads as failure. If the site only counts big wins, it will stop running the second, less photogenic event, which is where most of the money is.