The general adage is the more data you have, the better off you can be. The reasoning is that data is knowledge, and the more you have knowledge, the more you have to learn.
That is generally true up to an extent. If you have data but fail to annotate or filter the data, you have a mountain of stuff that could be good, but the work still has to be done to make it good.
It's almost easier to have a smaller amount of good stuff than to have a large amount of stuff that needs to be filtered for good. From an effort level it's more efficient. From a cost perspective it seems more efficient as well.
My new general adage seems to be - ensure that good quality is there first. You can do more with that data in the long run.