Abstract Introduction The aim of this study was to evaluate three-dimensional (3D) set-up errors and propose optimum margins for planning target volume (PTV) coverage in head and neck radiotherapy. Methods Thirty-five patients were included in the study. The total number of portal images studied was 632. Population systematic (Σ) and random (σ) errors for the patients with head and neck cancer were evaluated based on the portal images in the caudocranial longitudinal (CC) and left-right lateral (LR) direction measured in the anterior-posterior (AP) field, as well as from the images in the caudocranial longitudinal (CC) and dorsoventral lateral (DV) direction measured in the lateral (LAT) field. The values for the clinical-to-planning target volume (CTV-PTV) margins were calculated using ICRU Report 62 recommendations, along with Stroom's and van Herk's formulae. Results The standard deviations of systematic set-up errors (Σ) ranged from 1.51 to 1.93 mm while the standard deviations of random set-up (σ) errors fell in between 1.77 and 1.86 mm. The mean 3D vector length of displacement was 2.66 mm. PTV margins calculated according to ICRU, Stroom's and van Herk's models were comprised between 1.95 and 6.16 mm in the three acquisition directions. Discussion and conclusions Based on our results we can conclude that a 6-mm extension of CTV to PTV margin, as the lower limit, is enough to ensure that 90% of the patients treated for head and neck cancer will receive a minimum cumulative CTV dose greater than or equal to 95% of the prescribed dose.