From an Xoogler's perspective, the launches are all about the promos. Eg, just think about all those engineers who would not have gotten a promotion because then they would not have had a product launch in their promo packet.
And you have to be very conservative with what you consider a violation of that threshold.
Do scientific advances in papers also count as product launches, or do you have to put a thing in the world?
The current iteration has a 4-axis evaluation metric [1], where the value of a given project is measured by the area of the polygon.
So, you can make theoretical advances that don't go into products, but it's better for you if they do.
[0]: Disclaimer: I've been recruited by Google's research arm, but haven't worked there.