Why tolerances cost more than you think
A tolerance is a promise about how precisely a feature must be made, and precision is not free. Tighter limits force slower cutting feeds, finer tooling, more careful setup, and far more measurement. Each step adds labor and machine time that the part itself never shows.
The trap is that tolerances feel like a free number on a drawing. They are not. A team that over-specifies just to be safe can quietly double the cost of a simple bracket, because every feature inherits a tolerance it never needed.
How tolerance cost actually scales
Cost does not rise in a straight line with tightness; it climbs steeply. Moving from plus or minus 0.1 mm to plus or minus 0.05 mm might raise cost modestly, but the jump to plus or minus 0.01 mm can triple it, because the whole method changes, now you need a precision grinder, a temperature-controlled room, and CMM inspection.
Think in bands, not single numbers. A small tightening inside a band is cheap; crossing into a tighter band triggers new equipment and new process control. Knowing the band edges lets you stay just inside the cheap zone.
The practical tolerance bands
For CNC machining and most printing, plus or minus 0.1 to 0.2 mm is the default cheap zone where standard machines and normal feeds apply. With a little care, plus or minus 0.05 mm is reachable at modest extra cost.
Below plus or minus 0.02 mm you enter grinding, lapping, and CMM territory, where price climbs fast and lead time stretches. Reserve that range for the few features that truly demand it, never the whole part.
Tolerance only where features mate
Bore-and-shaft interfaces, sealing faces, and snap fits need tight values because they physically interact. Free surfaces, cosmetic edges, and internal ribs usually do not, and should sit at the generous default.
A part with one tight hole and ten loose faces should carry the tolerance on the hole only. Spreading tight tolerances across the whole drawing multiplies cost without improving function, the opposite of good tolerance design.
Tolerancing holes, shafts, and fits
Use standard fit systems, metric H7 g6 style limits and fits, rather than inventing arbitrary plus-minus numbers. Standardized holes and shafts let the shop select off-the-shelf drills, reamers, and gauges, and make assembly predictable.
Decide the fit by function: a clearance fit for free movement, a transition fit for locating, an interference fit for permanence. Specify the bore as the constant basic hole and vary the shaft, which is the cheaper, conventional practice.
GD&T versus plus-minus dimensions
Plus-minus tolerances on each dimension assume the part is measured from a corner that may not exist, and they let errors add up across features. GD&T controls the relationship of features to datums, which is usually what function actually requires.
Use position, flatness, perpendicularity, and concentricity to say what matters: that a hole sits where it should relative to a surface, not merely that two numbers are close. GD&T often lets you loosen individual dimensions while improving real fit.
Material and process effects on achievable tolerance
Plastics shift with moisture and temperature, so a tight plastic tolerance is partly illusion unless the part lives in a controlled environment. Size plastic tolerances generously and verify under the real use conditions.
Cast parts move more than machined ones, and 3D-printed parts vary with orientation and layer artifacts. Match the tolerance to the process's natural capability, not to an ideal number from a textbook, or you will pay for a tolerance the process cannot hold.
Datum and reference selection
Datums are the surfaces or axes everything is measured against, so choosing stable, machinable, and accessible features as primary datums keeps inspection repeatable. A datum that is hard to seat makes every downstream measurement noisy.
Pick datums that match how the part is actually located in assembly, the mating face, the bore, the mounting plane, so the tolerance you draw reflects the function you care about, not an arbitrary corner.
Stack-up and tolerance analysis
Individual tolerances accumulate along an assembly. Two parts each at plus or minus 0.1 mm can combine to plus or minus 0.2 mm at a critical gap, and three or four features in a chain can exceed the allowed clearance entirely.
Run a simple stack-up, worst-case or statistical, before releasing the drawing, especially for sliding, sealing, or alignment features. If the sum breaks the fit, loosen the non-critical tolerances rather than tightening every one.
A practical tolerance checklist
Before releasing: tight values only on mating features, standard fits for holes and shafts, GD&T for relationships, datums tied to real locating surfaces, and a stack-up check on any gap that matters.
Document the functional reason for every tight callout so the shop does not guess, and so future editors know which numbers are sacred. Most over-tolerance comes from fear, and a written reason replaces fear with evidence.
How FabProto supports tolerance design
FabProto returns a clear capability statement for your process, what tolerance band is standard, what costs extra, and where your drawing sits relative to the cheap zone. We flag over-tightened features before cutting begins.
Send one CAD model and we will advise the right tolerance strategy for your stage and material, including first-article inspection data so you can verify the fit. That feedback loop keeps your tolerances functional, not fearful.