The loss tree: naming where a sewing line's efficiency goes

A sewing line ran at 47 percent efficiency yesterday. Said alone, that number starts arguments and settles nothing. The supervisor blames the cutting supply. The IE blames the balance. The planner blames the style change. Everyone is partly right, and the meeting ends where it began.

The problem is not the number. The problem is that a single percentage hides the six different places the minutes actually went. The loss tree is the discipline of naming them.

The anatomy of a lost day

Line efficiency is earned minutes over attended minutes. Whatever the line earned, the gap to 100 percent is not one loss; it is a sum:

• Balancing loss. The line's bottleneck operation sets the pace; every other operator's spare seconds accumulate into idle capacity that was paid for and never used.
• Changeover. The hours between the last piece of one style and full speed on the next. On a floor with frequent style changes, this is routinely the largest single recoverable loss.
• Downtime. Machines that stopped, and how long repair actually took.
• Absenteeism. A line balanced for 28 operators does not degrade gracefully to 24. Denominators should reflect who attended; blame should not land on people who were never there.
• Rework. Minutes spent sewing the same seam twice, made visible only when rework time is captured where it happens.
• Method and everything else. The residue that remains when the named losses are subtracted, which is where genuine method study begins.

Why naming matters

Each loss has a different owner and a different fix. Balancing loss belongs to the IE and the skill matrix. Changeover belongs to method and sequence planning. Downtime belongs to maintenance. Absenteeism belongs to HR policy and rostering. Rework belongs to quality at the needle.

A single efficiency number lets the meeting argue about whose fault the day was. A loss tree assigns each slice to the person who can actually move it, with the minutes and the money attached. The argument becomes a queue of improvement work, ranked by what each loss costs.

Measured, not estimated

A loss tree built on assumptions is a story. Two rules keep it honest. First, unmeasured is not zero: if no hour recorded changeover minutes, the tree says changeover is unmeasured rather than pretending it did not happen. Second, improvements are measured twice, on the same basis and window, before and after. A kaizen without an equal-window remeasure is an anecdote.

What changes on the floor

The 07:30 meeting stops relitigating yesterday and starts assigning it. The line that read 47 percent decomposes into 9 points of balancing, 11 of changeover, 6 of downtime, 5 of absenteeism, 3 of rework, and the floor discusses the 11 first because it is the biggest and it has an owner. Improvement stops being a poster on the wall and becomes a register with baselines, remeasure dates, and held gains.

The efficiency number does not get kinder. It gets useful.

Further reading: Line Efficiency: Five Ways Factories Get It Wrong and Absenteeism Is an Hourly Number.