An error which is not random and is therefore liable to introduce bias into an estimator. For example, ${\displaystyle \sum _{r=1}^n\frac{{(x_i-\overline{x})}^2}{n}}$ has a tendency to underestimate the population variance. In a production process, if one of ten prongs on a rotating drum is broken then every tenth item on the production line will be faulty in addition to any random faults which may occur.