How tight is too tight? A real-world CNC tolerance guide for 2026

Specifying ±0.005 mm because "tighter is safer" can multiply your part cost by 5-10×. Here is what tolerance you actually need, by feature type.

Why tolerance is the #1 cost driver

Most prototype quotes we receive specify tighter tolerances than the part actually needs. The instinct makes sense — tighter = safer, right? Wrong. Cost grows non-linearly: ±0.1 mm is standard mill output, ±0.025 mm doubles the cost, ±0.01 mm requires grinding or careful CNC and quadruples cost, and anything tighter than that lives in lapping/EDM/jig-grinder territory at 8-15× standard.

The economics are simple: tighter tolerance = slower spindle speeds, more passes, in-process gauging, climate-controlled inspection, more rejected parts, more setup time. Every step multiplies. Yet most parts have one or two features that need precision, and twenty that don't.

How to actually choose tolerances

Map every feature to a function. Bearing bores? H7. Mating slip fits? H8. Cosmetic edge? ISO 2768-m (±0.1-0.3 mm). Threaded hole position? ±0.1 mm. Press-fit? Compute from the actual interference you need. Most general dimensions can simply default to ISO 2768 medium class and free up your supplier to use standard tooling.

Apply tolerance only where it solves a real problem — a fit, a function, a regulatory requirement. The rest defaults loose. We've seen customers cut $40 unit costs to $8 just by relaxing the "default" tolerance from ±0.025 to ±0.1 and tagging only the 3 critical features.

What we recommend in practice

Default ISO 2768-m on every drawing. Then call out specific features with explicit numbers — H7 bearing bore Ø25.0 +0.021/0.000, slip-fit shaft Ø24.95 ±0.01, threaded boss true position 0.1 mm. This is exactly how aerospace prime-tier suppliers spec their parts. It works for them, and at our scale it is what makes the difference between a $25,000 prototype run and a $4,000 one.