Purpose: First, to investigate the set-up improvements resulting from the introduction of a customised head and neck (HN) support system in combination with a technologist-driven off-line correction protocol in HN radiotherapy. Second. to define margins for planning target volume definition, accounting for systematic and random set-up uncertainties. Methods and materials: In 63 patients 498 treatment fractions were evaluated to develop and implement a 3D shrinking action level correction protocol. In the comparative Study two different HN-supports were compared: a flexible 'standard HN-support" and a 'customised HN-support'. For all three directions (x, y and z) random and systematic set-up deviations (1 S.D.) were measured. Results: The customised HN-support improves the patient positioning compared to the standard HN-support. The 1D systematic errors in the x, y and z directions were reduced from 2.2-2.3 mm to 1.2-2.0 mm (1 S.D.). The 1D random errors for the y and z directions were reduced from 1.6 and 1.6 mm to 1.1 and 1.0 mm (1 S.D.). The correction protocol reduced the 1D systematic errors further to 0.8-1.1 mm (1 S.D.) and all deviations in any direction were within 5 mm. Treatment time per measured fraction was increased from 10 to 13 min. The total time required per patient, for the complete correction procedure, was approximately 40 min. Conclusions: Portal imaging is a powerful tool in the evaluation of the department specific patient positioning procedures. The introduction of a comfortable customised HN-support, in combination with an electronic portal imaging device-based correction protocol, executed by technologists, led to an improvement of overall patient set-up. As a result, application of proposed recipes for CTV-PTV margins indicates that these can be reduced to 3-4 mm. (C) 2003 Elsevier Ireland Ltd. All rights reserved.