I liked this talk from Tim Harford from the Wired conference a while back. He uses some great examples to talk about the requirement for both types of innovation - that which brings incremental gain and continuous improvement over time, and the kind that involves high risk, different thinking, frequent failure but big potential gain.
The former can achieve remarkable things, and I think is naturally far more comfortable and palatable for large, established organisations. The latter is rare and typically avoided. But both, says Tim, are vital: "marginal improvement is all about climbing the hill you're on...longshots are all about finding new mountains to conquer". Absolutely.
HT @jamescaig for the link