Personal ultraviolet radiation (UVR) dosimetry has been performed for decades to objectively measure human exposure to UVR. These measurements have been used to investigate solar behaviour and its negative effects on human health such as skin cancer and positive effects such as vitamin D formation. A specific electronic dosimeter is described with a spectral sensitivity as the erythema response for human skin and temperature measurements for compliance control. Technical, methodological and environmental causes of uncertainties regarding personal UV dosimetry are investigated using this dosimeter as an example, which enables us to show the dosimeter's limitations and enables readers to compare their dosimeters with that described and to increase awareness of imperfections of dosimeters. The dosimeter's spectral response, cosine response, linearity, temperature dependency and sensitivity are investigated. As opposed to biological and chemical dosimeters, electronic dosimeters do not measure UV radiation continuously but at time-intervals (sampling). The error introduced by sampling is investigated for sampling intervals from 1 second up to 60 seconds for 3 groups of people (n = 18, 1.1-4.6 hours of positive UV measurements) on sunny (n = 12) and cloudy (n = 6) days. Increasing the sample time by 1 second added on average an uncertainty of maximum +0.29% to -0.27% per added second compared to the 1-second sample time. The importance of dirt on the sensor was investigated in 24 dosimeters after 6 months use by farmers. The reduction in the registered dose due to the dirty sensor was 2.3% (median = 2.0%, inter-quartile range = 2.0%, max = 5%) suggesting that dirt on the sensor generally does not play a significant role.