Shop-Floor Working Tolerance: Where to Split the Tolerance and Where Not To

Where the operation loses tolerance

The drawing sets the final size of the finished part. The shop, however, does not work with one number alone. It works through a chain of operations: rough machining, finishing, sometimes a re-clamp, a flip, or a second setup. That means the tolerance starts getting used up not at the end, but along the way.

The most common mistake looks reasonable only on paper: the drawing tolerance is split almost evenly between operations. The logic makes sense, but that calculation leaves out the things that really affect size on the shop floor: the first setup, tool wear, runout, part re-clamping, and measurement spread.

This is especially noticeable on a CNC turning shop floor. If the part is clamped once, machined, and removed, there is only one source of error. If it is then moved to another chuck or flipped, another deviation appears. Each one is small, but several of them can easily add up to a size going out of limit.

Usually the tolerance disappears in several places at once: during the first location of the part against the base, during a new setup, as the insert wears toward the end of the batch, and during intermediate inspection when the next pass has not yet been accounted for. That creates an unpleasant situation. After each operation the size seems fine, the part is still “in spec,” and it is passed on calmly. But every next step eats into the remaining margin. In the end, the finished size goes out of bounds even though in-process inspection never showed a clear problem.

That is where the shop-floor working tolerance becomes a trap. If you assign it without considering the accumulated errors, the route itself starts increasing the risk of scrap. This is especially dangerous on dimensions that close the tolerance chain. There is usually much less margin there than the drawing makes it seem.

The practical approach is simple: don’t split the entire drawing tolerance, but only the remaining allowance after setup, wear, and inspection have been taken into account. If the remainder is too small, it is better to rethink the machining route right away than to search for someone to blame after the batch has started.

Which dimensions can be split by operation

Not every dimension should be converted into a shop-floor working tolerance. The safest ones are the dimensions that are fully formed by a single finishing operation from the same base that is later used for inspection. If the base does not change, the machine is not rebuilding the size from scratch at every step, and the risk of error is much lower.

A good example is a diameter finished in one setup with a finishing pass. The same applies to the length of a shoulder or a face if they are produced in that same setup and checked immediately in the same way.

What matters is not the number on the drawing, but how the size is created on the machine. A dimension is suitable for splitting if the finishing operation forms it completely, roughing leaves a clear and repeatable stock allowance, the tool holds size steadily across the batch, and the shop and final inspection measure it in almost the same way.

It is especially convenient when roughing leaves a uniform stock allowance, for example 0.3–0.5 mm per side. Then the finishing pass removes nearly the same layer from each part and behaves predictably. If the stock allowance varies, the finishing pass starts behaving differently from part to part, and the calculation quickly loses its meaning.

There is also a simple practical sign. If the operator measures the first, tenth, and last part of the batch and gets almost the same result with the same micrometer or gauge, the dimension is usually suitable for splitting. If the result depends heavily on who is measuring, how the part is held, and how the measuring tool is applied, splitting the tolerance without a margin is risky.

On a CNC lathe, this usually works for outside diameters, simple fit lands, and lengths taken from a reliable face. If a shaft journal is left with a uniform roughing allowance, then one finishing pass brings the diameter in, and inspection confirms the same measurement with the same micrometer, the scheme is predictable. That is where you should start.

Which dimensions are better left alone

There are dimensions where even a small relaxation between operations later backfires, not at the machine, but at assembly. If a dimension affects how the assembly works, the shop-floor working tolerance must be calculated very carefully.

The first risk is tied to the tolerance chain. If a dimension closes a fit in assembly, you cannot split it just for the sake of machining convenience. Everything may look fine on a single operation, but in the assembled unit the deviations add up, and the part no longer fits as it should. This often happens with center distances, stop lengths, and fit diameters.

The second risk appears when a dimension is built from different bases. On paper the tolerance zone is one, but on the shop floor it is assembled from two independent steps. As a result, each operation stays within its own control limit, while the overall dimension drifts. This is a common trap in turning and milling: one side is machined from one base, the other side from another, and the total error becomes larger than planned.

The third case is thin walls and flexible areas. While the part is clamped, the dimension may look good. After unclamping, the wall springs back, and the actual value changes. If you also widen operation tolerances in such an area, the scrap risk rises immediately. You often see this on cups, bushings, thin-walled housings, and long parts with pockets.

Treat fits for a bearing, bushing, or shaft, sealing surfaces, diameters and faces that affect runout, and dimensions that influence coaxiality in the assembly as nearly untouchable. If the shop widens such zones without a calculation, the problem shows up too late: the part passes in-process inspection, but later causes noise, leaks, misalignment, or a tight assembly.

The rule is simple: if a dimension is tied not only to one operation, but also to locating, deformation, or assembly, do not rush to turn the drawing tolerance into a looser working one. First check how that dimension behaves in the finished part and what happens if it shifts even by a few hundredths.

How to convert a tolerance into a working one, step by step

A shop-floor working tolerance should not be taken from the drawing “by eye” and split evenly between operations. First you need to understand which dimension really affects the part after assembly. It may be a bearing fit diameter, a distance between bases, or a depth that determines how another part seats.

If you make such a dimension simply “easy for the machine,” scrap will increase even if the process sheet looks correct. That is why the calculation should start not from the operation, but from the function of the part.

1. Find the functional dimension on the drawing. Ask a simple question: if this dimension moves to the upper or lower limit, will the part still work or not? If it will not, then this is the dimension that sets the limits for all previous operations.
2. Identify the operation that fully closes this dimension. It is usually the finishing operation. The last operation must hold the zone confidently on its own, without counting on the next step to fix anything.
3. Leave a real margin for the shop. This includes tool wear, setup repeatability, chuck runout, and measurement error. If the size is checked with calipers and the zone is narrow, the machine spec alone is not enough.
4. Then set intermediate limits for each operation. Each operation needs not just one nominal value, but a permissible window: what can be passed on, and what cannot. It is also worth writing down the action to take when the limit is approached: change the insert, adjust the offset, check the setup, or stop the batch for measurement.

You can see it clearly on a simple part. Suppose the finished diameter must fall within 30.000–30.020 mm. If the finishing pass is really affected by 0.006 mm of tool wear, setup adds another 0.004 mm, and inspection adds 0.003 mm, the free range is smaller than it seemed at first. That means the maximum size after semi-finishing cannot be set “with an eyeballed margin.” It has to be shifted so that the finishing operation can still hold the drawing tolerance without strain.

That is how you get a shop-floor working tolerance: from the part function to the last operation, and then back along the route, with clear boundaries and a clear response from the supervisor and the operator.

What to check before calculating the range

It is risky to calculate the working tolerance too early. If the base, stock allowance, and measurement method do not match each other, the calculation will be neat only on paper. On the machine, it quickly turns into extra scrap.

First, figure out which base the operator uses for each operation. The drawing may show one base, but in the real setup the part is often seated against another surface, the jaws, or an arbor. Then the dimension may look the same, but it now lives in a different tolerance chain. The technologist splits the range by one logic, while the shop works by another.

It helps to walk through the part operation by operation and answer one question for each dimension: what is it actually measured from on the machine and during inspection? If on the turning operation the dimension is taken from a face after facing, but inspection is later done from the other face, a hidden error has already appeared.

Next, compare the real stock allowance with what is written in the process sheet. On paper there may be 2 mm per side, but after cutting the blank, sawing, or the previous operation, much less may remain. Then the tool removes a different amount of metal, the cutting conditions change, and the size starts drifting. On thin walls, this shows up especially fast.

Then assess what moves the size during machining. Clamping can crush a thin part or shift it off the base. Play in the axes and screw wear create different results when approaching from different directions. Heating of the part, chuck, and tool changes the size even within one batch. A long part can bend, even though everything looks fine on the inspection table.

The final check may seem boring, but this is often where the whole calculation breaks down. The gauge, micrometer, and machine program must all refer to the same point. If the operator measures the diameter in one zone, the setter adjusts by another, and the program calculates size from a different zero point, the shop-floor working tolerance loses its meaning.

In practice, a short check before starting the batch helps: where is program zero, from which surface does the inspector measure, where does the gauge touch, and at which point is the first trial size taken. Five minutes spent on that check often saves hours of sorting parts.

A simple part example

Let’s take a simple shaft. The drawing requires a diameter of 40.000–39.984 mm after finishing. The zone is small, only 0.016 mm. On such a size, the shop can easily make a mistake by splitting the tolerance equally between roughing and finishing.

On paper it looks neat. In real work it almost always works out worse.

The roughing operation on such a shaft should not hold half of the finished tolerance. Its job is different: leave a uniform and repeatable stock allowance for finishing, without large variation in ovality, runout, or tool marks. If roughing consistently leaves, for example, 0.20 mm on diameter with a spread of ±0.03 mm, the finishing pass works calmly.

Then the route looks like this: roughing brings the size to roughly 40.20 ±0.03 mm, and finishing removes the remainder and holds the finished size at 40.000–39.984 mm. The final zone is held by the last operation itself. Roughing does not have to hit the micron-level zone of the finished part. It only needs to bring the blank into a condition where the finishing tool can remove the stock steadily and achieve the required diameter.

Now look at the common mistake. The shop splits the final tolerance evenly and decides that after roughing it is already okay to aim almost at nominal, for example 39.992–40.000 mm, leaving the second half of the range for finishing. The problem is that finishing then has no margin for the real process: insert wear, elastic deflection, varying stock after roughing, and a little temperature drift.

As a result, finishing stops machining and starts chasing the size. On one part the tool removes enough, on another it removes less. The part quickly goes either undersize or stays oversize. It is not the machine expanding the scrap; it is the incorrectly assigned shop-floor working tolerance.

For a shaft, the rule is simple: if the final pass forms the size, do not split the drawing tolerance mechanically across the whole chain. First set a stable stock allowance after roughing, and only then give the finishing pass the full final-part range.

Where people most often create scrap themselves

Scrap often grows not because of the machine, but because the shop itself divides the tolerance zone poorly between operations. The mistake looks small on paper, but on a production run it quickly creates extra parts in the red zone.

The most common error is to take the drawing tolerance and simply divide it by the number of operations. Three steps means one third each. But the operations are not equal. Roughing controls size worse, finishing controls it better, and part of the variation comes from setup and tool wear anyway. If you split the zone without calculation, the shop-floor working tolerance is almost always false.

Another cause of scrap is a new base in the middle of the route. If the machinist or setter changes the base for clamping convenience, the logic of the related dimensions changes too. Then the dimension on the drawing seems to be under control, but coaxiality, runout, or distance to another surface drifts.

A separate problem is leaving too little stock before finishing. The shop wants to save time and removes almost everything earlier. Then the finishing pass no longer corrects anything; it becomes formal only. If the previous operation moved the size or shape, finishing will not pull the part back into tolerance.

Another common mistake is not revising working limits after a tool change or a new material batch. A new insert cuts differently. A different steel batch can behave differently in heat, chip formation, and size drift. If the old limits are kept, inspection is quiet at first, and then the shop gets a run of the same deviation.

Usually four checks are enough: calculate the real spread for each operation instead of splitting the zone evenly; keep one base logic across the whole route; leave stock that finishing can actually correct; and revisit the working limits after changing the tool or material.

In practice, most scrap appears not where the tolerance is narrow, but where the shop decided to simplify it without consequences.

A quick check before starting a batch

Before a batch starts, the shop often looks at the setup sheet but misses a simpler question: which operation actually forms the finished size? If finishing turning closes the size, do not comfort yourself with the idea that roughing was “almost fine.” Scrap is born in the operation that places the part into the final tolerance zone.

A quick check takes only a few minutes and often saves hours of sorting and rework.

• Assign the operation after which the size is considered finished.
• Check how much stock is really left before it, not by plan, but on the first part.
• Decide right away who measures the first and the tenth part, and with which tool.
• Set a rule in advance for what to do when the size moves toward the edge of the zone.

People miss the stock allowance more often than you’d think. If there is too little left before finishing, the tool can no longer remove size calmly after wear or a re-clamp. If the allowance is too large, cutting gets heavier, the part may spring back, and the size starts to wander. What matters is not the calculated remainder, but the actual one on the first blanks.

Measurement discipline matters too. The first part is usually measured by the operator right after machining, and then confirmed by the setter or shift supervisor. The tenth part should again be checked with the same agreed measuring tool. If the first was measured with a micrometer and the tenth with a different instrument and a different hand position, you get an argument, not control.

A useful rule is simple: the first and tenth part must be measured by the same method. The same bases, the same part temperature, the same tool. Otherwise you are comparing different results.

When the size drifts toward the upper or lower limit, the operator should not wait until the end of the ten-piece run. A clear reaction is needed:

1. Stop the run after the current part.
2. Recheck the size, tool wear, and setup.
3. Make a small correction and check the next part again.

The worst case is when the operator sees the size drifting but keeps producing parts as long as they still pass. That is how a tolerance zone quietly turns into a scrap zone.

What to do next on the shop floor

When the problematic dimension has already been found, do not leave the decision “in the head” of the supervisor or setter. Record it in the route sheet so that each operation has simple working limits: what can be held after roughing, what after semi-finishing, and what range is needed for finishing. The simpler the note, the less room there is for improvisation at the machine and during inspection.

If the dimension affects nearby bases or is part of a tolerance chain, do not write something vague like “keep it near the middle.” You need a specific range for each operation. Otherwise one person removes an extra 0.03 mm, another thinks there is still enough margin, and in the end you get scrap not from one mistake, but from the sum of several small shifts.

Before starting the full batch, it is useful to do a short check: machine the first parts using the new working tolerance scheme, show the result to the technologist and inspector right away, compare the actual dimensions with what was planned in the route, and adjust the operation limits if the real picture does not match the calculation.

This takes little time, but it often saves dozens of parts. Problems usually show up not in theory, but on the first blank: the tool drifts, stock allowance varies, and material from two batches behaves differently.

Another useful habit is to collect your real deviations. Do not rely on the phrase “we always machine it this way.” Pull measurements from several shifts: which size drifts up, which drifts down, where variation grows after insert changes, and where it grows after changeovers. After a week of notes like that, you learn more than from verbal rules.

Sometimes the problem is no longer in the tolerance calculation, but in the equipment itself. If the shop keeps running into repeatability, rigidity, or service-condition limits, one change to the route sheet will not solve it. In such cases, it helps to look not only at the cutting conditions, but also at the machine itself. EAST CNC has a blog with practical metalworking materials, as well as consultation, selection, commissioning, and service for CNC lathes. That helps when the shop wants not just to chase size, but to hold it steadily across a production run.