This paper describes the conceptual framework and the critical issues of investigations of clusters of childhood cancers and defines an investigative model for the health authorities responsible for assessing a suspected cluster, taking into account the guidelines available and considering the most recent advances of the Geographical Information System and of the specific statistical methodology. Three main investigation phases are identified: the first consists in the preliminary study on the health of population living in the area where the cases are defined and aetiological hypotheses are formulated; the second is the cluster evaluation study using statistical methods assessing the spatial heterogeneity and collecting information about potential risk factors; the third is the analytical epidemiological study to test aetiological hypotheses suggested by the previous phases. The residential cohort approach is the most valid to date to assess long-term effects, and allows to reconstruct the lifetime residential history from the population registry. The researchers' decision on how detailed about a suspected cluster the investigation has to be needs to take into account both the level of alarm in the population and the limited resources available. The concern about a suspected cluster of cancer cases should always be addressed, even if this implies to acknowledge limits of research and uncertainty in results interpretation.