This paper presents an ontology-driven workflow that feeds and queries a data warehouse opened on the Web. Data are extracted from data tables in Web documents. As web documents are very heterogeneous in nature, a key issue in this workflow is the ability to assess the reliability of retrieved data. We first recall the main steps of our method to annotate and query Web data tables driven by a domain ontology. Then we propose an original method to assess Web data table reliability from a set of criteria by the means of evidence theory. Finally, we show how we extend the workflow to integrate the reliability assessment step.