[petermcd]

I recall reading that back in the good old days at HP, they would split a group into two groups once it grew to over 100 people.

This discussion of Dunbar's number gives some useful background:

http://en.wikipedia.org/wiki/Dunbar%27s_number

[chalst]

When I was at KPMG, back in the early 90s, all the branches I visited had a similar structure - units of about 20 to 40 people, with an operational manager (OM, usually with senior managers as his/her title) who was not a partner. Each OM would report to at least one partner, but sometimes several OMs would report to one partner, or there would be a department with several units and several partners, where decisions would be made through a board of partners and OMs.

I never heard that this was a uniform policy for KPMG, but as far as I saw, it was how they organised audit, tax, and consulting units, and they used it for offices as small as 40 staff and offices as big as 700 staff. KPMG management back then were proud of how there was workplace harmony in the units, efficient communication between management and partners, and their federal structure (i.e., the top partner is not a dictator).

Groups of up to 40 staff would be around a third or a quarter of the optimal group size that Robin Dunbar posited for humans. Cf. http://blog.sitefox.com/wp-content/uploads/2010/06/Hill_Dunb... - it makes sense that the size of work groups in a large company should be below the Dunbar number, since their colleagues should not only be the people in their work group, but they should try to maintain a network throughout the company.

[simondlr]

It just makes a lot of sense. Interaction dynamics change at various sizes.