Ask HN: Are we collecting more data, more noise, or more rare events?

I posted this as a blog entry and in quora, but I would also like to hear any thoughts from the hn community on this.

We are in a data splurge. Everyone is interested in data, and developing data driven products. We collect tons of data about our users.

But how does one decide what data is worth collecting? And how do you strike the balance between collecting just increasing noise vs those events that will likely give us the crucial insight?