That's a great question.
I totally agree that too much data is confusing.
The real solution to that is not counting, and that's why it's not good as a metric to say your measure of outcome is x number of data sets released. That's not the right metric of outcomes for an open data site.
You can get around this and you can have lots of data out there, but what you need—and Ron touched on this—is a good search engine.
I spoke a bit about needing a better way to tag data sets in terms of issue areas. If you can organize your data that way, it doesn't matter. Look at StatsCan. They have tens of thousands of data sets, but they do it well so you can easily find what you're looking for by either subject area or term or search. I think those are what the government needs to work on—that search engine and that taxonomy—and those will go a long way. Then you can have 10 million data sets out there and it doesn't really matter.