Every drawing that leaves our studio carries dozens of tolerances. Most of them, in most studios, are inherited: copied from a previous revision, a supplier template or a habit formed years ago. Very few of them are questioned. That is a problem, because tolerance is where cost, lead time and reliability quietly collide.
A tolerance is a promise the machinist has to keep and the inspector has to prove. Every tight number on a drawing asks for a slower machine, a longer inspection and a higher chance that a good part gets scrapped because it missed a limit that never mattered. Multiply that by a few hundred parts and a few thousand units, and the most expensive decision in the project is often a number nobody remembers writing.
Precision is a budget
Tightening a bore from ±0.05 mm to ±0.01 mm does not just make a part more accurate. It changes the machine it can be made on, the fixture that holds it, the gauge that measures it and the number of parts that end up in the scrap bin. Lead times grow because fewer suppliers can hold the number. Costs grow because every part has to be checked.
We treat precision like any other budget. There is a fixed amount the function actually needs, and our job is to spend it where it buys performance and nowhere else. A drawing with ten tight tolerances in the right places is better than a drawing with a hundred tight tolerances everywhere, because the machinist can see what matters.
Start from the load path
Before we dimension anything, we trace how force moves through the assembly. Where does the load enter, which surfaces carry it and where does it leave? Interfaces that carry load or locate a moving axis get the tight numbers. Covers, brackets, cable guides and anything that only has to look right get generous ones.
It sounds obvious, but it changes the conversation. Instead of asking what tolerance a feature usually gets, we ask what happens to the machine if this feature is at the edge of its limit. If the answer is nothing, the tolerance opens up.
Stack-ups before sketches
For every critical function, such as the position of a probe tip or the preload on a bearing, we build a tolerance stack-up early, while the geometry is still cheap to change. The stack-up tells us which parts actually contribute to the error. Usually it is three or four features out of hundreds. Those features get our attention. The rest get standard grades.
Datums the shop floor can use
A tolerance is only as good as the way it is measured. We choose datums that a machinist can clamp and an inspector can touch, and we reuse them across parts in the same assembly. When the fixture, the machine and the gauge all reference the same three surfaces, the numbers on the drawing mean the same thing at every step.
What changes on the floor
On a recent six-axis inspection gantry, rethinking tolerances this way cut machining time for the frame parts by 31% and removed two inspection steps, with no loss in positional repeatability. The part did not get worse. The drawing just got honest.
The effects reach beyond the machine shop. Suppliers quote faster when a drawing is not full of tight numbers, because they can see the job is realistic. Assembly goes more smoothly because the parts that must fit precisely are the ones that were made precisely. And when something does go wrong, the drawing points straight at the features that matter.
How we review a drawing
Before any drawing is released, one engineer who did not draw it goes through every tolerance and asks the same three questions:
What breaks if this feature is at the edge of its limit?
How will this be measured, and by whom?
Is this number here because of the function, or because of a template?
Anything that cannot be answered is opened up or removed. It is a slow review the first few times. After that it becomes the way the team draws.
The honest drawing
A good drawing is a piece of communication. It tells the people who make and inspect the part where to spend their care. When every number is tight, nothing is important. When only the numbers that matter are tight, the drawing tells the truth about the machine, and the machine is cheaper, faster to build and just as precise.
