deviation

A deviation is the signed difference between an actual measurement and the target or reference value it should match. If a part is supposed to be 25.000 mm and you measure 25.015 mm, your deviation is plus 0.015 mm. If you measure 24.987 mm, your deviation is minus 0.013 mm. The sign matters because it tells you which direction you are wrong, not just by how much.

In manufacturing and quality control, deviations are the core language of process control. Machinists, tool setters, and inspectors live by them. A deviation tells you whether a dimension is oversized or undersized, whether a temperature is too high or too low, whether a surface is convex or concave. Without the sign, you lose half the information you need to correct the problem.

How deviation differs from tolerance

Tolerance is the allowed range: a part might have a tolerance of 25.000 plus or minus 0.050 mm. Deviation is the actual shortfall or overage from nominal. A part can be within tolerance but still show significant deviations. Tracking deviations over time reveals drift in your process, even if individual parts still pass. This is why statistical process control relies on deviation data, not just pass/fail results.

Deviations accumulate in stack-ups. When multiple tolerances add together through a sequence of operations or assemblies, you sum their deviations to predict the final result. A typical example: if five machined surfaces each have a positive deviation of 0.010 mm, the cumulative deviation can push a final assembly dimension out of spec even though each individual surface was acceptable.

The term also extends beyond dimension. In metrology you track deviation from flatness, roundness, perpendicularity, and runout. In process control, you monitor deviation from setpoint in temperature, pressure, or flow. In all cases, the sign tells the direction of error and the magnitude tells the severity. This is why deviation is recorded, graphed, and trended rather than simply discarded after a pass/fail check.