The cost of variability to society is, in general, far bigger than the cost of the variation itself. It can include, for instance, the guarantee and repair costs and the time clients have lost waiting for those repairs. Taguchi and Clausing (1990) has proposed the use of the idea of ‘loss function’ in order to measure the loss to society caused by deviations from a ‘target value’. According with their definition a distribution of frequency presenting low variability but contained within tolerance limits (process A - Figure 1) is much better than one centred at a target value and presenting high variability (process B - Figure 1). Indeed, when deviations from a target value present low variability (process A), the adjustments are more easy and likely to be successful.