One of the key challenges of Business Intelligence is to offer business users the opportunity of asking business questions in an intuitive manner, and receiving answers they can understand and trust. The problem is not about computing the right answers for a given query: it's to interpret—and if necessary guide—the user's actions in order to deduce their intent and make sure we answer the question they intended to ask in the first place. As engineers, scientists and developers, we tend to adopt authoring metaphors which we find intellectually elegant but will puzzle most users. Subqueries, for instance, are simply out of reach of most users, and can be replaced by incremental construction of query elements. Moreover, the lack of semantic depth in many tools — Excel to start with — can encourage users to perform mathematically expressible yet generally meaningless computations, such as an average of percentages. More generally, the subtlety of business questions is largely ignored, and BI tools tend to make a priori, undocumented decisions about the intended semantics of business questions—leading not to wrong, but to misleading answers. Through the exploration of a number of real-life examples, this presentation aims at identifying how we can at last adapt our query metaphors to a general audience.