You also take rider comments. Then take a PSD of the accelerometer data and correlate that data with the rider comments. Finding the road input transmissibility would allow you to quantitatively measure the attributes that riders like/dislike.

Deflection testing has it's place, but it's not a substitute for something like this.

To start life out, I'd stick an accelerometer on the seat post clamp and stem. You can't use an iphone. It's not sensitive enough and the filitering wouldn't be right. I'd use about a 50g accelerometer with a 100Hz low-pass filter and a sample rate of maybe 250-500 Hz.

Not that I've thought about it or anything.