Second Shift or Bigger Building: Finding the Real Constraint in a Fab Shop

The problem: The shop is behind and the floor looks crowded, so the choices on the table are a second shift or a bigger building, and nobody can say which machine or stage is actually holding the work up.

The solution: Measure where jobs wait rather than where they are worked, because the constraint is almost always one stage and expanding the rest of the shop buys nothing.

The pressure arrives as lead time. Quoted at four weeks, shipping at seven. Customers are calling. The floor is full of steel waiting on something, the parking lot has a trailer being used as storage, and every conversation about it lands on the same two options: run a second shift, or add on to the building.

Both are big commitments and both assume the same thing, which is that the shop as a whole is out of capacity. That assumption is almost never true. Shops do not run out of capacity evenly. One stage runs out, everything upstream of it piles up in front of it, and the pile is what makes the building look small.

The floor lies to you about capacity

Walk a fab shop that is behind and you will see work in process everywhere. That is genuinely what an owner sees, and the conclusion that there is not enough room is a reasonable reading of the evidence.

But work in process is not distributed randomly. It accumulates in front of whichever stage cannot keep up. If the brake is the constraint, you will find laser-cut parts stacked around it, and you will also find them stacked wherever there is floor space near it, which is most of the shop. The visual signature of one overloaded machine is a crowded building.

This is why the space decision so often disappoints. A shop adds 15,000 square feet, the work in process spreads out into it, the floor looks better, and lead time does not move, because the same stage is still setting the pace. The addition bought organization, not throughput.

The second shift version fails differently. Running a full second shift across every department adds cost across every department, when only one of them needed the hours. It usually works, in the sense that lead time improves, and it works at several times the cost of adding hours only where they were needed.

What you would need to know

The question is narrow and answerable: which stage sets the pace, and by how much?

Four measures get you there, and none require new machinery.

Queue time by stage. How long a job waits in front of each stage before work begins on it. The constraint is the stage with the longest and most consistent queue. This is the single most useful number in the shop and almost nobody has it.

Machine utilization split into run, setup, and idle. A brake that is busy 90 percent of the time but spends half of that on setup has a changeover problem, not a capacity problem, and the fix is scheduling and tooling rather than a second machine.

Job routing versus actual path. Which jobs touch which stages. A constraint that only affects 40 percent of jobs is a different problem than one every job passes through.

Rework loops. Parts coming back through a stage a second time consume constraint capacity twice, and are usually recorded as ordinary work.

Get those and the decision generally makes itself. If one stage carries a queue several times longer than everything else, you have a machine and staffing question, not a real estate question.

Why the shop does not have these numbers

Not for lack of trying. Job travelers get stamped or scanned at completion, so the record shows when a stage finished a part and nothing about when the part arrived. Queue time is the difference between those two moments, and the second one is never captured.

The scheduling board, whether it is software or a whiteboard, holds the plan rather than the actual. The machine monitoring on the newer equipment, if there is any, reports its own uptime to its own screen and does not know what job it was running. Setup time exists in the estimator's standards and is not compared to what setups really take. Rework goes back through the shop on the original traveler.

So each stage's story is recorded somewhere, and no two of those stories are in the same place, which means the one comparison that would settle the expansion question cannot be made. The owner is left with the crowded floor, which is real evidence pointing at the wrong conclusion.

How the answer becomes available

The capture change is small: stamp arrival at a stage as well as completion. That one addition turns the traveler into a record of queue time across the whole shop, and it costs a scan.

From there, the job's routing, the estimated hours, the actual run and setup, the machine data, and the rework flag need to sit together in one connected picture, so a job can be asked where it spent its time. Once they do, the constraint stops being a matter of opinion between the shop manager and the owner and becomes a number that everyone can see.

Then the upkeep runs on its own. Rather than someone building a capacity study before a capital decision, automation watches queue time as it accumulates and flags where the pace is being set: the stage whose queue has grown for three straight weeks, the setup times drifting past standard on a particular machine, the job type that loops back through the constraint most often. Constraints move as the mix changes, which is exactly why a one-time study is the wrong instrument and a standing measurement is the right one.

A look at a metal fabrication shop

Consider a fabrication shop doing about $10 million a year, roughly 55 people, running laser, brake, weld, finish, and assembly in a 40,000 square foot building. Quoted lead time is four weeks, actual is closer to seven, and the owner is weighing a 15,000 square foot addition at roughly $150 a square foot, about $2.25 million, against a second shift estimated at around $900k a year in wages, premiums, supervision, and utilities.

Suppose that before deciding, the shop starts stamping arrival as well as completion for one quarter, and connects those stamps to routing and to the estimate. What you would expect to surface is not that the shop is out of space. It is a queue profile: laser and weld running with short, stable queues, assembly waiting mostly on parts rather than on floor, and one stage, plausibly the brake, sitting on a queue several times longer than anything else with utilization heavy on setup.

That reframes the money entirely. A second brake with tooling at roughly $280k and an operator at about $60k loaded is around $340k, and if setup analysis suggests a share of the constraint's lost hours are recoverable through better sequencing and tooling changes, a portion of the gain costs nothing at all. Against $2.25 million of building or $900k a year of second shift, the comparison is not close.

The likely conclusion an owner would reach is to relieve the constraint first, then re-measure. Lead time would be expected to improve as the queue in front of that stage drains, and the floor would look less crowded without a square foot being added, because the pile was the symptom. If the constraint then moves to weld, which it often does, the shop is in a good position, because it now has a standing measurement and can see where it moved. The building may still be justified eventually, on growth rather than on congestion, and by then the owner would have a record of what the last capacity investment actually returned to price the next one against.

How to start

You can do this in one quarter without buying equipment.

1. Stamp arrival, not just completion. One extra scan at each stage converts your travelers into a record of where work waits.
2. Rank the stages by queue time. The longest, most consistent queue is your constraint. Everything else is noise until that one moves.
3. Split the constraint's hours into run, setup, and idle. If setup dominates, the cheapest capacity you will ever buy is sequencing and tooling, not steel or square footage.
4. Let the measurement stand. Set automation to track queue time and flag where the pace is being set, so when the constraint moves you find out in weeks rather than at the next capital decision.

The takeaway

A crowded floor and a long lead time are real, and they are not evidence that the building is too small. They are evidence that one stage is setting the pace and everything upstream is stacking up in front of it. Find that stage by measuring where jobs wait, not where they get worked on, and the choice between a second shift, an addition, and a single machine usually resolves itself, often at a fraction of what the first two would cost. Buy space when you have proven you need space. Until then you may be paying seven figures to store the consequences of one bottleneck.