Before we can talk about getting exact answers from any B.I. tool, ad hoc or otherwise, we must discuss Data Quality.

Data quality starts with good data modeling and good database design.

To learn about how to "bake in" data quality in databases, we are publishing articles and white papers on the subject.

Our first article on Data Quality discusses one of the problems plaguing database modeling and design today: the prevalent use of Surrogate Keys. NOTE: see this article at SqlServerCenter: here

Google and other search tools scan unstructured documents and provide links to these documents. But their keyword algorithms don't provide exact answers, which are nearly always facts that live in structured databases.

Ad hoc query tools' great benefit over regular Search tools in their ability to produce "actionable facts".

End users need and want immediate access to actionable information stored in relational databases. To be actionable, the information and facts retrieved from an organization's data sources must be accurate.

While there are new ad hoc query tools emerging that generate wonderful looking graphs and dashboards, they offer no ability to 'certify' that the answers they report are correct (particularly if they generate queries against databases with many tables).

Ad hoc query tools are generally offered in two categories: Natural Language (EasyAsk *, Semantra *) and Visual Query tools (Visual Query, QlikView, Tableau).

Note: the author of iTrieve believes that a successful semantic search product must incorporate BOTH a Visual Query component and a Natural Language component.
This belief goes back to his first Natural Language invention, in 1984: SALVO. See a writeup of SALVO in the AFIPS '84 Proceedings, National Computer Conference, ACM New York: here