Shop floor management: the column with the longest dwell is your constraint

Every shop owner has a theory about where the week goes. It is usually a person, a supplier, or a stage that feels slow. The theory is usually wrong, and it is wrong in a specific and predictable way.

Loud is not slow

The stages people blame are the ones they can hear and see. The bay with the most movement. The stage with the most people in it. The one where the arguments happen.

Those stages feel like the bottleneck because they are where the attention is. But attention follows activity, and a bottleneck is usually the opposite of activity — it is where things sit and nothing happens to them.

A car waiting on a part makes no noise. A car waiting on an approval makes no noise. A car waiting for the one machine that can do the next step makes no noise. Silence is what a constraint sounds like.

The number that settles it

If work moves through named stages and cards move as the work moves, then every job carries a record of how long it spent in every stage. Average that per stage and you have dwell.

The stage with the largest dwell is your constraint. Not the busiest, not the loudest, not the one with the most complaints attached. The one where time accumulates.

This is almost always a surprise the first time. In mechanical shops it is very often waiting-on-approval — an entirely invisible queue made of cars sitting on unanswered texts. In collision it is usually one of the parts waits, but not the one anyone expected. In tire shops it is the rack.

Why it has to be split to be visible

Dwell only works if the stages are real. A board with three columns — new, in progress, done — will tell you that dwell is concentrated in "in progress," which you already knew and which helps nobody.

The resolution of the answer is the resolution of the board. If parts waits and active work share a column, you cannot tell a supply problem from a labour problem. If approval and estimating share a column, you cannot tell whether the delay is yours or the customer's.

This is the practical reason for splitting stages that feel similar. Not tidiness — measurability. Each split is a question you become able to answer.

What to do once you know

A constraint has three honest responses, and one dishonest one.

Add capacity to it. More of whatever the constraint is: another rack, another person on approvals, a second supplier.

Reduce demand on it. Route work around it where possible. Not everything that goes through the constraint has to.

Protect it. Make sure the constraint is never idle waiting for input. This is the cheapest of the three and the most commonly skipped: a bottleneck that goes quiet for twenty minutes between jobs has lost throughput the whole shop pays for.

The dishonest response is to speed up a stage that is not the constraint. It feels productive, it is measurable, and it changes nothing about output. Work arrives at the bottleneck faster and waits longer. The pile just moves.

And then it moves

Fix a constraint and it does not disappear. It relocates. Whatever was second becomes first.

That is not failure, it is the process working. It does mean shop floor management is not a project with an end. The useful habit is checking dwell periodically — monthly is plenty for most shops — and asking the same question again: where is time accumulating now.

The shops that get durably faster are not the ones that found the answer once. They are the ones that kept asking, and stopped guessing.