My boss and I spend a certain amount of time discussing how to measure what we do and determine success, a particularly difficult problem for a design organization. We aren't, after all, able to conduct experiments like having two teams in isolation working on the same problem, one with design process and deliverables, the other with none. Sutton advises that if you can't do a controlled experiment, at least make sure your supposed cause precedes your supposed effect :-)
Sutton also comments on a few "evidence-based" writers on management and innovation:
I also love Larry Prusak, the knowledge management guy. He is smart, well-read, humble, opinionated, and evidence-based. And much better read in academics than any academic I know. ...
I also like Malcolm Gladwell and Steve Levitt and their models of evidence-based management. Although, I think that Blink is much weaker than The Tipping Point because Gladwell misses that the instant judgments he praises are developed through years of experience. You have to do a lot and think a lot about something before you can do the blink thing.
Finally, a prediction, I think the next big guru, at least if ideas and ability to present them counts, is Chip Heath of the Stanford Graduate School of Business. He has a book coming out with his brother called What Sticks about the kinds of ideas that people remember and what persists. It not only is based on sound research, it is also hugely practical (I’ve seen executives get so excited by his stuff that they immediately grab him for gigs).
Design management is a special kind of management, and requires special kinds of innovation grounded in both evidence and insight. I've been reading and enjoying Managing as Design, another Stanford Business product. It's a great collection of fairly academic and thoughtful pieces on organizations and management especially around design, with a fair amount of thought on architecture in it.
Edited to add: I forgot I meant to talk about evidence in management and the role of conversation and email in decision-making. Scott Berkun reminded me in his post on The Things You Never Hear. There's all sorts of decision-making in organizations that's based on slim or biased reporting from the "ground" and managers often never hear -- or don't want to hear -- the things that would make them more effective. Worse, in a social networking sense, there's diagonal or peer-level information that would be useful, that the relationships or physical proximity or office politics don't support hearing. Gossip can be important in an office, but it does need to be evenly spread to be of a useful evidential nature. There's another whole post in this somewhere, though...