We're worried about that too. What we're hoping to do as we progress is to use best practices as metrics. Because we can do semantic matching, and not just textual matching, there's a world of possibilities to make results more helpful