Tolerance Stack-Up Analysis: A Complete Guide for Mechanical Engineers

You send a part to manufacturing. Three weeks later it comes back. It doesn't fit.

The drawing looked fine. The model looked fine. The individual tolerances were all within spec. But when you assembled everything together, the cumulative error pushed the final dimension outside the acceptable range.

This is a tolerance stack-up problem — and it's one of the most common and expensive mistakes in mechanical engineering.

This guide explains what tolerance stack-up analysis is, the two main methods for doing it, and how to work through it step by step.

What Is Tolerance Stack-Up Analysis?

Tolerance stack-up analysis (also called tolerance chain analysis or dimensional chain analysis) is a technique for calculating how individual dimensional tolerances accumulate across an assembly.

Every manufactured part has tolerances — small acceptable deviations from the nominal dimension. A shaft might be specified at 25.00 ± 0.05 mm. A bore might be 25.10 ± 0.05 mm. Individually, both parts are within spec. But the clearance between them depends on both tolerances simultaneously. If the shaft comes in at the high end and the bore comes in at the low end, the clearance is tighter than expected. If this happens across multiple parts in a stack, the final gap — or interference — can fall outside the functional requirement.

Tolerance stack-up analysis quantifies this risk before any parts are made.

Why It Matters

The cost of catching a tolerance problem depends entirely on when you catch it:

• In design review: a few hours of engineer time
• After first article inspection: a drawing revision, maybe a new prototype run
• After parts are manufactured: potential scrap, rework, or redesign
• After the product is in the field: warranty claims, recalls, reputation damage

Most tolerance problems are entirely predictable. The reason they make it to manufacturing is that the analysis wasn't done — usually because it's slow, tedious, and easy to defer.

The Two Main Methods

1. Worst-Case Analysis

Worst-case analysis assumes every dimension in the stack simultaneously hits its worst possible value. It gives you the absolute limits: if every part is at the extreme end of its tolerance, what is the resulting gap or interference?

Formula:

For a linear stack of n dimensions, each with tolerance tᵢ:

Total tolerance (worst-case) = Σ |tᵢ|

Where each tᵢ is the individual tolerance (half the tolerance band if bilateral).

Example:

Suppose you have a housing with 3 components in a linear stack:

| Component | Nominal | Tolerance (±) | |-----------|---------|---------------| | Part A | 50.00 mm | ±0.10 mm | | Part B | 30.00 mm | ±0.05 mm | | Part C | 20.00 mm | ±0.08 mm |

Total nominal = 100.00 mm Worst-case total tolerance = 0.10 + 0.05 + 0.08 = ±0.23 mm

This means the final dimension can range from 99.77 mm to 100.23 mm.

When to use worst-case:

• Safety-critical applications where you cannot accept any failures
• Small assemblies with few components in the stack
• When all parts are manufactured by the same process with predictable variation
• When the cost of a single failure is very high

Limitation: Worst-case analysis is conservative. The probability of every part simultaneously hitting its worst-case value is very low in real production. For large stacks, this can lead to unnecessarily tight tolerances and higher manufacturing costs.

2. RSS Analysis (Root Sum Square / Statistical)

RSS analysis treats tolerances statistically. Instead of assuming every part hits its worst case simultaneously, it assumes part dimensions follow a normal distribution (which they typically do in controlled manufacturing). The resulting assembly variation is the square root of the sum of squared individual tolerances.

Formula:

Total tolerance (RSS) = √(Σ tᵢ²)

Using the same example:

RSS tolerance = √(0.10² + 0.05² + 0.08²)
             = √(0.0100 + 0.0025 + 0.0064)
             = √0.0189
             ≈ ±0.137 mm

The RSS result (±0.137 mm) is significantly less conservative than the worst-case result (±0.23 mm). This means you can use wider individual tolerances while still achieving the same assembly performance — lowering manufacturing cost.

When to use RSS:

• Large assemblies with many components in the stack
• High-volume production where statistical assumptions hold
• When some failure rate is acceptable and the risk can be quantified
• Consumer products where manufacturing cost is a key constraint

Limitation: RSS is only valid when part dimensions are normally distributed and processes are in statistical control. It also implies a non-zero reject rate at the assembly level. Standard RSS assumes 3-sigma limits, which gives approximately 99.73% yield.

Step-by-Step: How to Perform a Tolerance Stack-Up Analysis

Step 1: Define the functional requirement

Start with the question you're trying to answer. This is usually a gap, clearance, or interference that must stay within limits.

Example: "The axial clearance between the gear and the housing wall must be between 0.1 mm and 0.5 mm."

Step 2: Identify the dimension loop

Draw a closed loop that includes every dimension contributing to the functional requirement. Every dimension in the path between your two reference surfaces belongs in the stack.

A common mistake is missing a contributing dimension — for example, forgetting that a fastener head sits on a countersink, adding another variable to the stack.

Step 3: Assign directions

Each dimension in the loop either adds to or subtracts from the total. Define a positive direction (typically left-to-right or top-to-bottom) and assign +1 or -1 to each dimension accordingly.

Step 4: Calculate nominal and tolerance

Sum the nominal values (with signs) to get the nominal result. Then apply your chosen method (worst-case or RSS) to calculate the total tolerance.

Step 5: Compare to functional limits

If the nominal ± total tolerance falls within your functional requirement, the design works. If not, you need to either tighten individual tolerances or redesign the assembly.

Step 6: Optimise

When the stack fails, don't just tighten everything. Use the analysis to identify the biggest contributors (the terms with the largest tᵢ values) and focus tolerance tightening there. Tightening a tolerance that contributes only 5% to the total variance has almost no effect.

Common Mistakes

1. Doing it too late. Stack-up analysis is most valuable in early design, when changing a dimension costs nothing. Engineers often defer it until after detailed design is complete, when changes are expensive.

2. Forgetting GD&T contributors. Flatness, perpendicularity, and runout tolerances all contribute to a stack-up and are frequently omitted from manual calculations.

3. Using worst-case when RSS is appropriate. Over-constraining a design leads to unnecessarily high manufacturing costs. Use the method that matches your production environment and risk tolerance.

4. Not updating the analysis when the design changes. The stack-up is only valid for the design it was run on. If dimensions change during development, the analysis needs to be rerun.

5. Analysing in 1D when the problem is 3D. Linear stack-ups are the most common, but some assemblies require 2D or 3D analysis, especially when angular or radial dimensions are involved.

Tolerance Stack-Up in Practice

In most engineering teams, tolerance stack-up analysis is done in Excel. This works — but Excel-based analyses have real limitations: they're time-consuming to set up, easy to get wrong, hard to review, and they don't update automatically when the design changes.

If your team is doing regular stack-up analysis, it's worth having a structured process and dedicated tooling so that analysis happens earlier (when it's most useful) and more consistently.

ForgePilot includes a tolerance stack-up analysis tool that lets you define a dimension chain, input your tolerances, and get worst-case and RSS results immediately — without setting up a spreadsheet from scratch each time.

Summary

Tolerance stack-up analysis tells you whether your design will assemble correctly in production. The two main methods — worst-case and RSS — give you different confidence levels and suit different applications.

The key is doing the analysis early, covering all the contributing dimensions, and revisiting it whenever the design changes. A stack-up problem caught in design review costs hours. The same problem caught after manufacturing costs weeks.