The prognostic impact of different anatomical sites in patients with cutaneous malignant melanoma (CMM) has been widely debated and requires further elucidation. Therefore, we developed EssDoll, a new computerized method to address the question of site in relation to prognosis. A population-based cohort of 1891 patients, diagnosed between 1976 and 1987 with invasive CMM without evidence of metastasis, was identified. The body surface was divided into 24 areas. Hazard ratios (HRs) for CMM death were calculated and areas were compared in both the whole model and in pairs. Cox's proportional hazard regression model was used and adjustments were made for established prognostic factors. Furthermore, the overall effect of site was calculated using the likelihood ratio test. Overall, the tumour site was of prognostic importance (P=0.0036). There was a significantly increased risk of CMM-specific death in patients with a primary tumour site in the middle and lower back (HR=1.8, P=0.04) and in the supramammary and mammary area (HR=1.8, P=0.05). When all areas were analysed in pairs, the dorsal shoulder, superior back and clavicular area also showed a worse prognosis. CMM diagnosed in other anatomical regions, including the calves, Achilles, upper arms, forehead, temples, cheeks and face, seemed to be related to a better prognosis. It can be concluded that the tumour site is of prognostic importance, and that the middle and lower back and supramammary and mammary areas are independent factors related to a poor prognosis.