Database schemas are used to describe the logical design of a database. Diverse groups of users have different views on the global schema which leads to different local schemas. Research has focused on view integration to generate a global, consistent schema out of different local schemas. However, this approach is too constrained when the generated global view should be variable and only a certain subset is needed. Variable schemas are needed in software product lines in which products are tailored to the needs of stakeholders. We argue that traditional modeling techniques are not sufficient for expressing variable database schemas. We show that software product line methodologies, when applied to the database schemas, overcome existing limitations and are able to generate tailor-made database schemas.