Personal data anonymization requires complex algorithms aiming at avoiding disclosure risk without compromising data utility. In this paper, we describe a model-driven approach guiding the data owner during the anonymization process. Depending on the step, the guidance is informative or suggestive. It helps in choosing the most relevant algorithm given the data characteristics and the future usage of anonymized data. It also helps in defining the best input values for the chosen algorithm. The contribution is twofold: a meta-model describing the anonymization process and components and an approach based on this meta-model. In this paper, we focus on microdata generalization algorithms. Both theoretical and experimental knowledge regarding anonymization is stored in an ontology. An experiment, conducted with sixteen participants allowing us to check the usability of the approach, is described.