← Notes ·
The 22-minute changeover that takes 95
A worked example of SMED on a real bottling line, and the cadence trap that almost killed the savings before they landed.
The line was supposed to changeover in 22 minutes. The work-study said so. The operator manual said so. The capacity model that justified the £4M line extension assumed it.
The actual changeover, measured cold over three weeks, was 95 minutes. On a line running three changeovers a shift, two shifts, six days a week, that’s 1,300 minutes of lost output a week. At the line’s £1,800/hour contribution, that’s a £39,000-a-week hole nobody on the operations team had named.
The SMED walk
Two of us spent a Tuesday with stopwatches. Not the formal kind — just the iPhone clock, recording each step the operator did between “last bottle of A” and “first bottle of B” passing the inspect camera.
What we found:
- 8 minutes of genuine internal work (cap-feed swap, label-magazine swap, height-rod reset)
- 14 minutes of external work being done internally (fetching the changeover parts trolley from the prep area, three trips)
- 22 minutes of internal work that should have been one operation but was two people walking past each other (line clearance + label reload colliding at the inspect station)
- 31 minutes of waiting (waiting for the lab to clear the rinse, waiting for QC to release the new batch, waiting for the warehouse to deliver the empty pallets)
That’s 75 minutes accounted for. The other 20 was small stuff — a stuck label cassette twice, a printer that needed a hard reset, two operators looking at the same SOP page.
The fix nobody wanted
The fix wasn’t a SMED training session. The fix was a changeover cadence. Three rules:
- Parts trolley arrives at the line ten minutes before line stops. Done by the team leader, not the operator. (Kills the 14 minutes of fetching.)
- QC release happens during the previous batch’s final 30 minutes. Not after the changeover starts. (Kills the wait.)
- One person does internal work, one person does external. They swap when the line restarts. No more collisions at the inspect station. (Kills the 22 minutes of step-on-each-other.)
After three weeks, the new mean was 34 minutes. Not 22 — there were still genuine internal steps, and the lab couldn’t always release early. But 34 instead of 95 is £25k/week recovered. £1.3M/year. From a Tuesday with a stopwatch.
The trap that almost killed the saving
Three months later, the new mean was back up to 67 minutes.
The reason: the team leader got pulled into a project, and “parts trolley arrives ten minutes early” stopped happening. Within four weeks, the old pattern reasserted itself completely.
We re-installed the cadence — same three rules, same names against them, plus a Tuesday morning audit on the changeover-prep step. Six months later it was holding at 38 minutes.
The lesson: SMED is the analysis, but cadence is what holds the saving in place. Skip the cadence, and the analysis was free consulting that nobody bought.
The SMED Analysis tab on the £97 Toolkit is the same template we used for the stopwatch walk. The cadence rules go on the DMS Huddle pack.