When you're not shipping, you're not learning from users. As a result, it's easy to build working, correct, performant code which doesn't fit what anyone actually needs.
The trick is they just complain about the last thing they remember being bad, so it's a good sign when that doesn't change, and it's bad if they start complaining about a new thing.
> Figuring out what is useful for people is not some difficult problem that requires shipping half baked slop
what have you shipped? paying sees literally hundreds of thousands of dollars a year to ship out fledged out software that no one wants is exactly why Stadia lasted both way too long and got cancelled anyway.
figuring out what is useful is the hardest problem. if anything that's Google's biggest problem, not shipping fast enough, not iterating fast enough.