Jeff Veen has a good post describing his working method in approaching design and how it has a lot to do with the inspiration that occurs after stewing in lots of collected data:
This leads me to believe that doing research in web design -- for me at least -- has more to do with Method Acting than ethnography. Robert De Niro used this technique as he prepared for his roll as Travis Bickle in Taxi Driver, spending a month pulling late night shifts as a cab driver. He did this not to mimic those in the profession, but to be able to react on screen in a way that they would. Applied back to design: Rather than figure out how to design for your audience, design for yourself after becoming like your audience. At that point, I find, snap decisions become good decisions.
The piece is a summary/continuation of his How to Inform Design talk at SXSW, which struck a nerve with me when I heard it because I'd always felt a little strange (insecure is probably a better word) working in exactly the way Jeff describes. Everyone else, with their mountains of data, carefully crafted use case scenarios, iron-clad five step design processes, personas, and strategic analysis made my off-the-cuff process feel a little inadequate. Several years ago, I figured out that a significant part of a designer's job was to (somehow) come up with the correct solution for a given problem and then sell that decision to the client. For me, like Jeff, the solution would usually arrive fairly late in the game after I'd been soaking in the problem for 90% of the available time. But like he says, you can't sell client services with that approach, which is why you need to sell all the fancy sounding stuff and just get it done right however you can.