Over the years, companies like Oracle have tried to develop data-mining applications for use by non-programmers. The goal is to give middle-management a way to work with data-views as objects, combining them, interesecting them, taking the intersection set and treating it as a new view and then doing further intersections with it, etc etc.
The major limitation of these attempts, IMO, has been to assign each view to a page in a sequence of pages. The metaphor was too sequential and not spatial enough. Other approaches use ERD diagrams and lines representing joins. That metaphor is too spatial and not sequential enough, since it typically displays only one query at a time, and it is difficult to "see" the relationship between and among a set of queries.
But if we could have a large workarea with Euler circle diagrams that can be intersected and unioned with other Euler circles, and if the intersection could be "perforated" so it could be torn off and dropped onto another "blank" Euler circle to become its dataProvider, or two or more Euler circles could be encompassed in another to become its unioned dataProviders, then data could be manipulated visually, and you'd have both a spatial and sequential view of the data. If the workarea could scroll, and present the data manipulation stages in vertical tiers, one wouldn't have to leave the page to go back to an earlier stage, but could simply move up a tier by scrolling. You could "follow the data mining narrative" more easily.