Data warehouses increasingly play important roles in the information technology landscape of the financial industry. However, semantic heterogeneity is high in banking – data is defined differently by different banks, business units, and users. Therefore data integration in financial data warehouse development projects relies on the knowledge, know-how, and judgment of human experts. Up to now, methodical support is missing for the communication process among experts that determine and negotiate a shared understanding of requirements. In contrast to ontologydriven or schema-matching approaches proposing the automatic resolution of differences ex-post, we introduce an approach that addresses data integration already in early project phases. Our approach supports developing shared understanding of domain concepts and data fields in financial data warehouse projects, good communication of all participants while the project progresses, and early detection of errors within projects. This way, we prevent problems that result from the ex-post resolution of semantic heterogeneity.