How Are Tolerances Determined on Engineering Drawings?
The Essence of Tolerances: Why We Need Them
In an ideal world, every part would be made exactly to its nominal dimensions — but in real manufacturing, that’s impossible.
Machines vibrate, tools wear, materials expand or shrink, and humans make minor errors.
That’s why tolerances exist:
“A tolerance defines how much a part is allowed to deviate from its nominal geometry while still functioning as intended.”
Without tolerances, two mating parts might not fit, a shaft might seize in its bearing, or a cover might rattle in assembly.
So, tolerances are not arbitrary — they are a functional, economic, and technical compromise between ideal design and real-world manufacturing.
What Makes a “Good” Tolerance Design?
A common misconception is:
“The tighter the tolerance, the higher the quality.”
Not exactly.
Tighter tolerances do increase precision, but they also increase cost, lead time, and difficulty.
A hole controlled to ±0.005 mm might sound impressive — but it could require high-end CNCs, temperature-controlled environments, specialized inspection tools, and a much slower machining speed.
That’s why good tolerance design balances three things:
- Functionality — The part must perform its intended job.
- Manufacturability — It must be possible to make the part repeatedly and efficiently.
- Economy — It must not drive unnecessary cost or complexity.
In other words:
“The best tolerance is not the tightest one, but the one that’s just tight enough to make the part work reliably.”
Step-by-Step Guide: How to Define Tolerances Logically
A robust tolerance design process typically involves five major steps:
Step 1: Understand the Function and Critical Features
Before you even start assigning numbers, ask:
- What is this part supposed to do?
- What are the critical features that affect that function?
For example:
- Shaft and hole fit → control diameter tolerance.
- Sealing surfaces → control flatness and surface roughness.
- Bolt pattern alignment → control position tolerance.
- Guiding mechanisms → control parallelism or straightness.
If you don’t fully understand the function, you can’t design the right tolerance — it’s like prescribing medicine without a diagnosis.
Step 2: Rank the Functional Priorities
Not all dimensions are created equal.
Some features directly affect performance or safety, while others are secondary or cosmetic.
You can roughly divide tolerances into three levels:
| Level | Function Type | Example | Control Type |
| Primary | Critical to fit, function, or safety | Shaft/hole fits, bearing bores, sealing faces | Tight tolerance or geometric tolerance |
| Secondary | Affects performance or assembly ease | Mounting hole patterns, alignment slots | Moderate tolerance |
| Tertiary | Non-functional, aesthetic, or reference features | Non-contact surfaces, chamfers, general outlines | General tolerance per ISO 2768 |
This hierarchy ensures you spend your precision budget where it actually matters.
Step 3: Define Reference Datums — The “Coordinate System” of Your Part
In GD&T (Geometric Dimensioning and Tolerancing), datums are the reference points, planes, or axes from which all measurements are taken.
They ensure consistency across design, manufacturing, and inspection.
When selecting datums, follow these principles:
- Functional principle: Match how the part is positioned in real assembly.
- Stability principle: Choose stable, repeatable, and rigid features.
- Inspectability: Make sure the datum can be accessed for measurement.
- Simplicity: Use as few as necessary — typically 1 to 3 are enough.
A well-chosen datum structure can reduce measurement ambiguity and production errors dramatically.
Step 4: Choose the Right Type of Tolerance Control
There are two main families of tolerances:
- Dimensional Tolerances — Control size variation (e.g., ±0.02 mm).
- Geometric Tolerances (GD&T) — Control form, orientation, or location.
| Type | Controls | Examples |
| Form | Shape | Straightness, flatness, roundness |
| Orientation | Tilt or alignment | Perpendicularity, parallelism, angularity |
| Location | Relative position | Position, concentricity, symmetry |
| Runout | Combined motion error | Circular or total runout |
Rule of thumb:
If a simple dimensional tolerance cannot guarantee assembly or performance — use geometric tolerances.
For instance, two holes might both be Ø10 ±0.02 mm, but if their position drifts by 0.5 mm, the assembly still fails. That’s why you specify position tolerance.
Step 5: Determine the Numerical Tolerance Values
Finally, assign actual numbers based on three key factors:
- Functional Requirements
Use engineering calculations or simulation to determine how much variation the function can tolerate.
Example: For a press-fit, you might calculate necessary interference based on load and material yield strength.
For alignment holes, position tolerance may be derived from bolt diameter and clearance.
- Standards and References
Refer to tolerance tables such as ISO 286 (for fits), ISO 2768 (for general tolerances), or company-specific design guidelines.
These standards summarize achievable precision based on manufacturing methods.
- Process Capability
Even if theory allows tight tolerance, the factory might not.
Check your suppliers’ machining and inspection capability — e.g.,
CNC milling (IT7–IT9 range)
Grinding (IT5–IT6)
Laser cutting (±0.1 mm typical)
Designing beyond real process capability just causes rework, cost overruns, and frustration for everyone involved.
A Practical Way to Think About It
Next time you assign a tolerance, ask yourself three questions:
- What function does this feature serve?
- Can it be made and measured easily?
- If it’s off by X, does it still work?
If the answer to the last question is yes, your tolerance might be too tight.
If the answer is no, it might need refinement — or better functional analysis.
Final Thoughts
Good tolerance design is both science and art.
It requires understanding the physics of how parts work, but also the pragmatism of how they’re made.
Tolerances connect the digital world of CAD with the real world of manufacturing.
Mastering them means fewer production problems, better quality, and happier suppliers.
As the old saying in precision engineering goes:
“Anyone can design a perfect part — only a good engineer can design a manufacturable one.”











