Encyclopedia · Free preview
Tolerance Stacking in Assembly
Tolerance stacking is the phenomenon where small dimensional variations in individual parts compound when assembled, potentially causing the final assembly to fail specification even though each part passes inspection. If ten parts each have plus-or-minus 0.5mm tolerance assembled end-to-end, total tolerance is plus-or-minus 5mm. Machinists manage this through statistical analysis (root-sum-square rather than worst-case arithmetic), datum referencing (measuring from the same surface), and selective assembly (matching oversize with undersize partners). Toyota's obsession with reducing variation at each step is fundamentally about preventing tolerance stacking. System-level quality is not the sum of component quality—it's constrained by how variations compound through the chain.
When to use it
When multiple sequential steps or components must work together within defined specifications. When troubleshooting quality problems that appear at the system level despite acceptable individual components. When designing processes where you must decide whether to tighten individual tolerances or reduce the number of tolerance-contributing steps.
How it can help
Tolerance stacking is one of the most practically important models for anyone managing multi-step processes. In software, small bugs in individual modules compound into system failures. In project management, small delays across sequential tasks create massive deadline overruns. In communication chains (telephone game), small interpretation errors at each step produce nonsensical outputs. The model teaches that controlling variation at each step is exponentially more valuable than fixing problems at final assembly. It also teaches the power of reducing the number of steps in a tolerance chain—simplifying the assembly is often more effective than tightening individual tolerances.
Keep exploring
Read the full page.
Create your free access to continue reading and explore the complete library.
Register free with ChatGPT →Already registered? Use the same button to sign in.
Sign-in shares your email with Michael Simmons to create your site access. No payment required. Newsletter signup is separate. How your data is used