My good friend Love just provided me with a paper by Cook, Grange, and Eyre-Walker. About group-size and productivity. They have played around with group-sizes, different types of group-members, and different types of output. And they have some good data to work with, and do good statistics. A good paper it seems to me.
Part of the group at a writing-session in the north
To start with, the output does increase with the size of the group, but not proportionally. It is not one output-situation where y=x. And while post-docs and Ph.D. students seems to increase productivity more than technicians, post-docs seems to be the only parameter that increases impact and citations. The size of the group seems to have little to do with this, impact and citations.
Part of the group visiting an excavation in Kiev
I guess if you really wanted to shape your work after the data, a few independent medium-sized or small sized groups mainly with post-docs would be the best. Then the next question is of course, why are we doing this and what is it that we think is the fun in it. For me it is the questions and materials I work with, and the people I collaborate with. It may not be the best approach for shaping the perfect group, but it is good enough, and it is the best approach for me.
We are allowed to some non-scientific fun every now and then