Design AtlasNo manufacturing process ever produces a dimension exactly as drawn — injection molding shrinks by a slightly different amount lot to lot as it cools, machining tools wear, assembly jigs sit a hair off position. So a drawing always carries both a nominal value and an allowed deviation, written like "⌀20 ±0.05" — meaning the part passes anywhere from 19.95 to 20.05mm.
Tighten the tolerance (say, ±0.01) and mating parts fit precisely with no play, but the process and inspection needed to hit that get sharply more expensive. Loosen it (say, ±0.2) and manufacturing gets easy and cheap, but stacking several such parts together compounds their individual tolerances — "tolerance stack-up" — and the final assembly risks being loose or simply not fitting. In practice, only the dimensions that functionally need precision get a tight tolerance; everything else gets as loose a tolerance as the design can tolerate, to protect cost.
In fit terms, the worst-case (tightest) combination is the smallest hole paired with the largest pin; the loosest combination is the largest hole with the smallest pin. A designer has to verify the part still works at both extremes — tolerance design isn't drawing one number, it's designing the whole range that number allows.
When to use
When designing two mating parts (a hole and a pin, a threaded fit), or reviewing a dimensional chain across several stacked parts.