Engineering Reference

Tolerance Stack-Up Calculator

Combine up to six tolerances by the worst-case and root-sum-square methods, and see how much the statistical approach saves.

Data verified 2026-09-29 · based on n/a — standard engineering relationships, no single governing revision

Quick Answer

Worst case adds the tolerances: ±(t₁ + t₂ + … + tₙ). The statistical (RSS) method combines them as ±√(t₁² + t₂² + … + tₙ²). With ten equal tolerances, RSS gives ±3.16t against a worst case of ±10t — a 68% reduction.

Tolerance Stack-Up

The Formulas Used

Worst case (arithmetic): T = Σ tᵢ
Statistical (root-sum-square): T = √(Σ tᵢ²)
For n equal tolerances: worst case n·t, RSS t·√n

Worst Case vs Statistical — and When Each Is Right

The worst-case method assumes every dimension lands at its limit simultaneously, in the direction that makes the assembly fail. It is guaranteed correct — if the parts pass inspection, the assembly works — but it is expensive, because it demands tight tolerances on every contributor regardless of how many there are.

The RSS method assumes the errors are independent and normally distributed, so they partially cancel. It is statistically valid for a large production run and lets tolerances be much looser. But it carries a real risk: a small percentage of assemblies will be out of tolerance, and the method is only valid if the processes are actually centred and in statistical control.

The choice is a business decision, not a mathematical one. Worst case is right where a single failure is unacceptable — safety-critical assemblies, aerospace, medical. RSS is right for high-volume commercial production where a small reject rate is cheaper than tight tolerances on every part. A common compromise is a shifted RSS using 1.5 times the RSS value, which approximates a process that has drifted off centre.

Frequently Asked Questions

What is the difference between worst-case and RSS tolerance stack-up?
Worst case adds all tolerances arithmetically and assumes they all land at their limits in the same direction. RSS combines them as the square root of the sum of squares, assuming independent normally distributed errors that partly cancel. RSS gives a much smaller total but accepts a small statistical risk.
When should I use worst-case tolerance analysis?
When a single out-of-tolerance assembly is unacceptable — safety-critical parts, aerospace, medical devices, or any assembly where the consequence justifies the cost. Worst case guarantees that any combination of conforming parts will assemble.
How much does RSS save?
For n equal tolerances, worst case is n·t and RSS is t·√n, so the reduction is a factor of √n. Ten tolerances give RSS at 32% of worst case; four tolerances give 50%. The saving grows with the number of contributors, which is why RSS is most valuable on long stacks.
What are the assumptions behind RSS?
That each dimension is independent, normally distributed and centred on nominal, and that the processes are in statistical control. If a process drifts off centre, or if dimensions are correlated — as they are when several features are machined in one setup — RSS understates the real variation.
What is the 1.5 sigma shift?
A convention from Six Sigma practice that assumes a process mean drifts up to 1.5 standard deviations off centre over time. Applying it widens the RSS estimate substantially and is a more realistic model of production than a perfectly centred process.

Related

Value Sources

Each data column on this page is tied to the source it came from. The numbers in square brackets correspond to the table headers above.

#SourceTypeRevision / method
[1]ASME B1.1 — Unified Inch Screw ThreadsstandardASME B1.1-2019 — source
[2]ASTM A615 — Deformed steel bars for concrete reinforcementstandardASTM A615/A615M-20 — source
[3]ASTM E140 — Hardness Conversion TablesstandardASTM E140-12b — source
[4]Values computed in your browserderivedEvaluated locally from the formulas shown on the page. No data leaves the device.
[5]ISO 4287 — Surface texture: Profile methodstandardISO 4287:1997 — source
[6]ISO 68-1 — Basic profilestandardISO 68-1:2023 — source
[7]NFPA 70 NEC Table 310.16standardNEC 2023 (NFPA 70-2023) — source

Data Sources

StandardRevisionWhat it covers on this page
Formulas as shown on this pagen/a — standard engineering relationships, no single governing revisionevery value this calculator produces

Cross-checked against:

Derived values — the following values on this page are calculated, not taken directly from the standard:

ValueHow it is derived
All outputsComputed in the browser from the formulas above. No data leaves the device.

Outputs are computed from the formulas shown. Verify against the governing standard for design or acceptance work.

Accuracy and use. The values on this page are compiled from the published standards and cross-checked sources listed above. Where values are derived, the derivation is stated. No warranty, express or implied, is made as to the accuracy or completeness of this information, and no liability is accepted for any loss or damage arising from its use. Engineering reference data is provided for guidance in preliminary work — before a value is used for design, fabrication or acceptance testing, verify it against the current revision of the governing standard and against your own inspection. The user assumes all risk and responsibility in connection with the use of this information.

Cite This Page

Every value on this page is traceable to the sources listed above. If you use the data in a document, paper or report, cite it as:

Each row in the tables above also has a permanent link — hover a row and use the # link to cite a single value rather than the whole page.