An assembly's real tolerance is a chain, not any single part's number
A multi-part assembly's actual fit depends on every dimension in the chain between two reference points, not just the one part you happen to be looking at. If three components each contribute their own small tolerance to the same gap, the total uncertainty at that gap is the sum (or, statistically, a combination) of all three, not just the largest one alone. Skipping this step is a common way an assembly that looks fine on paper, part by part, turns out not to fit once everything is stacked together.
How the stack-up works
Each component's tolerance is entered and summed for a worst-case total, the conservative assumption that every component could land at its extreme simultaneously. The tool also identifies which single component contributes the largest share of that total, since tightening the dominant contributor reduces the overall stack-up more than spreading the same effort thinly across every component.
Worked example
Bracket hole position ±0.15 mm, shaft diameter ±0.1 mm, standoff length ±0.3 mm: the worst-case total is ±0.55 mm, with the standoff length identified as the dominant contributor at 55% of the total, meaning a tighter or better-measured standoff length would do more for the overall assembly fit than tightening either of the other two components.
A common mistake
A frequent error is stacking tolerances from parts that were never actually measured, treating a designer's nominal tolerance as if it were confirmed real-world performance. A stack-up is only as trustworthy as its weakest input; a guessed tolerance for even one component undermines confidence in the whole total, so measure the dominant contributor directly whenever possible.
Limitations
This sums the component tolerances you enter transparently; it does not know the true tolerance of any component you have not actually measured, and worst-case summing is a planning tool, not a statistical guarantee of assembly success.