Of Tolerance Worksheet

How to Build a Practical Tolerance Stack-Up Worksheet

Tolerance stack-up analysis is one of those things that sounds straightforward until you actually try to close an assembly and everything binds. I built a lot of spreadsheets early in my career before settling on something that actually survives contact with machinists and QA engineers who want the raw data, not a summary.

What a Real Of Tolerance Worksheet Looks Like

A solid worksheet tracks every individual dimension in a stack, assigns a tolerance to each, and then calculates the cumulative effect on the critical gap or interference you're trying to control. Most people start with the wrong columns. They put tolerance on its own line instead of treating it as a property of each part. That makes the math harder to follow and makes it easy to lose track of which tolerance is which when the stack gets long. Here is the structure that works in practice.

Direction is simply positive or negative along your chain. A washer stacking up adds in one direction; a gap subtracts. Writing this out explicitly prevents the classic error of treating every value as additive. I once spent two days debugging a stack where a machined face was labeled correctly but the tolerance direction was implicitly negative due to the part's orientation. The result was an interference of 0.18mm that never showed up in the spreadsheet. I caught it by walking through each contribution on paper and comparing it to the actual assembly sequence. Now I add a quick sketch column referencing the drawing view number. It takes an extra thirty seconds per line and saves hours of confusion later.

The Two Methods You Actually Need to Know

There are two main approaches: worst-case and statistical. Every junior engineer picks one and treats it like the answer. That is a mistake.

Worst-Case Analysis

You add all tolerances algebraically in the same direction. The formula is simple. You take the sum of the absolute tolerances for the stack. If your gap has five contributors each at ±0.10mm, the worst case is ±0.50mm. Use this when:

Non-replaceable assemblies where replacement means pulling the whole product apart
Low production volumes where statistical sampling is meaningless
Regulatory or safety requirements demand zero risk of interference

The downside is that worst-case is brutal on design. Parts end up with tight tolerances that cost more to produce and do not need to be that tight in reality.

RSS (Root Sum Square) Analysis

This assumes tolerances follow a normal distribution and independent variations. You square each tolerance, sum them, then take the square root. For the same five contributors at ±0.10mm, RSS gives you roughly ±0.22mm instead of ±0.50mm. Use this when:

High-volume production with stable processes
Process capability is documented and Cpk is above 1.33
Multiple independent sources of variation exist

RSS underestimates when parts are not truly independent. I have seen couples of features machined from the same setup move together. Treating them as independent gives a falsely optimistic stack. The workaround is grouping correlated features into a single combined tolerance before feeding them into the RSS calculation.

Building the Worksheet Step by Step

Start from the datum. Every stack needs a clean reference. If your assembly drawing does not specify the datum clearly, ask before you build anything. I learned this the hard way on a housing assembly where two mating surfaces had different datum targets depending on who drew the detail view. The stack changed completely once I aligned the datums to the functional assembly sequence. 1. List every part in order from the fixed datum to the closing feature. 2. Assign a nominal value and tolerance to each. Pull from the drawing, not from memory. 3. Determine direction. Sketch a free-body chain. Mark each contribution as positive or negative. 4. Calculate worst-case by summing signed tolerances. 5. Calculate RSS by squaring, summing, and taking the square root. 6. Compare the result against the functional requirement. 7. Flag any single contributor that dominates the stack. If one part accounts for more than 50 percent of the total variation, tightening another tolerance will not move the needle. Fix the dominant dimension first.

Common Pitfalls That Waste Time

Forgotten floaters. These are parts that appear in the assembly but do not constrain the stack directly. A spacer that slides freely inside a bore is a floater. Including it as a fixed contributor inflates the stack. Exclude it, or add it as a separate contribution with its own direction logic. Ignoring thermal expansion. Sheet metal enclosures and aluminum internals expand differently. At temperature extremes, the stack shifts. I add a delta column for temperature sensitivity when the assembly spans more than a 20-degree operating range. It rarely changes the initial call, but it catches issues during qualification testing. Treating soft material deformation as rigid. Rubber seals, gaskets, and compliant surfaces compress. Their effective thickness changes under preload. Use measured compression curves from the supplier rather than nominal thickness. The difference between using a published curve and a nominal value can be 0.05 to 0.15mm on a seal stack.

When the Worksheet Lies to You

A stack-up analysis is only as good as the data entered. It cannot predict assembly error caused by operator technique, tool deflection, or fixture play. If your design relies on the analyzer to close the gap every time, you are designing around a spreadsheet. Build test fixtures and measure actual assemblies early. One physical build usually reveals what the worksheet missed within a day. If your stack has more than eight contributors and includes mixed manufacturing processes, consider Monte Carlo simulation instead of manual RSS. The worksheet becomes unwieldy, and the assumptions behind RSS break down under complex coupling. A simple @RISK or even a basic Python script with random sampling will give you a distribution shape that is far more useful than a single RSS number.

Practical Template for Quick Use

Set up your Of Tolerance Worksheet with these columns and nothing more:

Keep the source column. It forces you to cite where each number came from and makes revision tracking trivial. When a drawing updates, you change one line and the stack recalculates. Without the source, you spend twenty minutes chasing which tolerance belongs to which revision. I usually cap the stack at ten lines before splitting into sub-stacks. Anything longer tends to hide errors, and I stop trusting my own arithmetic. Sub-stacks that feed into a master calculation are easy to verify independently. The method is not perfect. It does not replace physical proof. But it catches the majority of assembly failures before tooling starts, and that is where the real value sits.