Finding the slowest station on a sewing line

Every operations manager knows the rule: a production line runs at the pace of its slowest station. Everyone nods. Few factories can say which station that is today.

Why a sewing factory is harder to read than a stamping plant

I spent twenty years at SNOP, an automotive Tier 1 that stamps and welds metal parts. There, the equipment was the data source: presses reported cycles, stops and utilisation without anyone asking. Managing meant reading what the machines already said.

I now lead FFG Bags in Vietnam, which makes technical bags and backpacks for outdoor, defence and industrial brands. The work is done by people on sewing machines. The machines are inexpensive and say nothing. Output changes with the operator, the day and the style. Large lines move together, so one slow station slows everyone, and nobody can see it by walking past.

A modest start: a proxy, not a measure

We could not measure the thing, so we measured something close to it. The team designed a simple, low-cost sensor for each machine. It picks up a signal that is a very good proxy for how intensely that machine is used. The electronic parts cost very little. Once we had defined what we needed, an AI assistant helped us build and configure them in-house.

We are careful about one point and repeat it on the floor: this is not a measure of productivity or efficiency. It is a hint. Its value is to open eyes.

How to read a proxy signal

A short checklist we use, and that applies to any imperfect metric:

1. Look for extremes first. The busiest and the quietest machines on a line are where questions begin.
2. Busiest machine: possible bottleneck. If it is busy because everyone else waits for it, the line’s pace is set there.
3. Quietest machine: look at its position. If it is the last in the line, the constraint is probably upstream, and the quiet station is simply waiting.
4. Odd pattern: suspect the machine. Sometimes an unusual signal is just a machine asking for maintenance.
5. Never conclude from the data alone. Take the finding to the floor and ask the people who run the line.
6. Say what the number is not, to the team and in every report.

This is the theory of constraints in plain form: improve the station that limits the line, and improving anything else changes nothing.

What the team showed us

A pilot on a few machines worked well. It quickly showed large differences that were worth checking with production staff, which is exactly the job of a proxy.

The next step is to compare each machine’s running time with actual output, so that we can see where to look first across days and stations: line balancing, training, maintenance or material supply. It is a pilot, run with the workers informed, and we say so.

The MBA lesson behind it

At the Executive MBA with ESSEC and Mannheim, I kept coming back to one idea: when you cannot measure the thing, measure something close to it, say clearly what it is not, and then go and look with your own eyes. A cheap first version, tested early and corrected, teaches more than a perfect system that arrives late.

A question to keep

Most operations run on numbers that are known to be imperfect. The danger is not using them. It is forgetting what they are not.