The stratification of salons with regard to workload levels was done to evenly distribute workload in the randomisation process. A possibility of bias could have existed if the subjects in the intervention groups had significantly different pre-intervention workload, since the chance of an intervention effect would be higher if the Intervention II group had a higher pre-intervention workload compared to the Intervention I group. The two intervention groups were also similar in background variables, although the Intervention group I had numerically (but not statistically significantly) more neck and shoulder complaints.