I’m going to throw out another hypothesis here:
I claim there are Team Size Ranks, numbers which are the average thresholds at which group dynamics fundamentally alter by orders of magnitude. It’s well known that communication because more difficult as you add people to a team because there are NxN communication paths in any group. It’s sort of roughly understood that things seem to roughly change too at certain thresholds – at some point you need a team manager, etc. What’s less understood is that there are key trip points where teams need to explicitly morph into something new or they will thrash and suffer. I claim these trip points are roughly at sizes greater then 3, 8, 25, 50 and 100.
In my experience, these thresholds radically alter the roles on the team. If you go up a level, all of a sudden, everyone has a different job. Being a producer means a different thing then before, being a lead means a different thing then before, being a designer means interfacing with X more people, even being a coder is different! A “lead” of 8 can still write or design whole systems. They can lead by example. A “lead” of 25 can’t afford that any more. They don’t have time. A “lead” of 50 probably can’t remember everyone’s name, let alone delegate, and probably needs “sub-leads”. And most of the time people aren’t trained or ready for that switch. Their identity is still caught up in the “lead” of 8, and they’ll fight to preserve that. Identity is the hardest thing for people to change. It’s what they wanted! It’s why they took this job! It’s their calling!
As teams grow, individuals become partnerships, then tribes, then tribes of tribes, then tribes of tribes of tribes. And each requires a radically different kind of management. With 100 people, producers have to be very careful with their impact, because the 100×100 matrix means their input will propagate far. At their best, producers shield and stop problems before they even hit the 100 person level – limiting them to a smaller scale.
I’m very aware of this dynamic on teams. These numbers are scary to me. Very scary. Teams make great games. And breaking these size barriers seems to be the biggest threat to them. Like throwing them into a giant whirlpool they may never come out of. It’s extraordinarily dangerous to throw 50 people on a team, even if they were a team before, because you jump a size barrier here. All of sudden, every person on that team is trying to understand a new role, something that can take a year or more to stabilize again, often with high attrition.
In fact, I’ll claim people can only go up (or down) 2 levels comfortably, and then start to really flounder beyond that. At that point, it stops being the job they wanted, and they’ll want to retrain or change jobs.
Some other interesting side effects:
As you go up, relationships require different skills and more specific compartmentalization. Each individual has to be more specialized, more focused, on a 100 person team. It’s much harder to be a generalist. You can’t want to touch everything – it requires too many communications to do that.
As a consequence, people (and work) done best with cross-discipline Face to Face interaction are happiest with small teams, and vice versa. Agile can help here, but you’re always sacrificing communications somewhere for the sake of others.
As you go up in size ranks, relationships become less human. Face to Face is a 3, maybe 7 person thing. Doesn’t work anywhere near as well with 100 people. This can be really disconcerting for people, particularly if they were used to solving problems on a smaller team. In 100 person teams, decisions start to become very impersonal, because they have to. The nature of the information management requires abstracting it out. Yet if you’re used to dealing with things directly, this can be very confusing.
I don’t know what’s the next ranks are. I’m pretty sure there’s more steps, starting around 200. Large corporations have thousands of employees and exhibit all the characteristics of large scale, yet even more extreme. My guess is there’s some empirical pattern here, but I haven’t been able to derive it on my own.