[enginous]

You're right, office conditions are notoriously hard to research because variables are abundant. However, there are ways to measure it empirically. In the design of an experiment like that, you'd want to use metrics that are heavily dependent on the change. Measuring financial outcomes is no good, because they are affected in significant ways by a number of other factors (including a large factor of luck).

A better metric for this kind of research could be stress levels because they are significantly dependent on changes in physical working environments. Stress is affected by a number of other work factors, such as management, work/life balance and workload, but these can be measured and controlled for. The effects of stress on various performance outcomes are fairly well understood, so this relationship can be used in tandem with other variables to ensure that a change is having the hypothesized effect. Again, it's not simple, but it's far removed from having no comparison.

I wouldn't suggest using online communications for every decision, but rather recording the essence of each substantial conversation digitally. Tasking one person with sending an e-mail containing the conclusion of a discussion takes a fraction of the man-hours consumed by a conversation, especially if many people are frequently involved (which I still advocate against).

Further, summarizing by e-mail is an aid to people's memories, helps avoid miscommunication by establishing a mutual understanding, and chronicles decisions and progress for later review or lookup.