>No, you often can't. If you could, it would be already done by one of those shallow initiatives the GP is talking about.
Well, a lot of them have. Besides "you often can't" is not very helpful.
The question is whether it's more difficult because you started as shallow or not -- not whether it's not a 100% surefire way to make it.
I'd be very interested on any ideas about how to make deep innovation less risky. For shallow ones is "launch fast", but what to do when this one is not available?