“Two boys is half a boy, and three boys is no boys at all”
Team size is such a significant driver of the structure and so much research in organisational behaviour, sociology and psychology has been conducted that it rapidly demanded its own post. Here, then, is that post.
Jeff Bezos, of the Amazon, is famously (if unreliably) credited with coining the two-pizza rule. This rule states that any team should be well-fed with only two pizzas, and is intended to imply that small teams are more effective than larger ones.
As someone who is both a little obsessed with getting the facts straight, and who has at one gluttonous point in my life halved a Domino’s XXL pizza with a colleague in a single sitting, I needed to research how big a team Jeff was implying with his pizza order. Voila!
Using the power of Domino’s Group Ordering Tool, we can reliably establish that the maximum size of a Bezoan team is 8 people. That’s real data for you, people.
So, we have now incontrovertibly proved that Jeff’s rule for team size is ‘up to 8’. This rule of thumb jibes nicely with a significant body of research into team dynamics which stretches all the way back to the early 20th century that suggests strongly that small teams are more effective.
But, why are small teams better?
Some of the earliest research into the size of effective teams was carried out by Maximilian Ringlemann, a French professor and agricultural engineer who devised an experiment to test the pulling force of individuals when performing alone and in groups.
Max asked his volunteers to pull a rope, on their own and in a group to which he added more and more individuals. As the group grew Max observed a decrease in the average force of the individuals (ie, each person pulled relatively less than on their own). Max’s experiment wasn’t constructed to determine whether the phenomenon, which is delightfully dubbed ‘Social Loafing’ (or the Ringlemann Effect), was due to a loss of motivation when individuals worked in a team or to a lack of communication between team members.
Later studies, by Alan Ingham in 1974 and Bib Latane in 1979, more clearly demonstrated that communication alone couldn’t cause the deterioration in personal effort, with Bib claiming:
“If the individual inputs are not identifiable the person may work less hard. Thus if the person is dividing up the work to be performed or the amount of reward he expects to receive, he will work less hard in groups”
Broadly, misanthropes among us might suggest that it’s easier to shirk responsibility as the group grows. Those of a more humane persuasion may argue that it’s simply harder for people to value their own contribution and motivate themselves in larger groups. Either way, numerous experiments backup Ringlemann’s work – bigger groups seem to exhibit worse individual performance.
Dunbar’s Number, Groups and Guilds
More recent research, notably that of Robin Dunbar, suggests that it’s not just a motive issue, but also one of our own human capacity to work effectively with groups.
Dunbar, a British anthropologist and evolutionary psychologist, proposed that a correlation could be seen between primate brain size and the number of members in their social groups. By extrapolating for human brain size, Dunbar suggested that human groups would have a comfortable limit of between 150 and 250 members. Research into sizes of early human groups, such as neolithic communities, seem to bear out some truth to this size with around 150 now being the popular value of what has come to be known as Dunbar’s Number.
Dunbar also further refined his number, with varying sizes of groups dependent on ‘closeness’, suggesting that the groups would diminish in size and increase in dependence roughly with the ‘rule of three’ – 150 being the number of people in a loose social group (say casual friends), 50 in a group of close friends, 15 in a close-knit group with which a bereavement may be shared, and 5 as a close support network. Clearly this varies by individual – for me as an unrepentant introvert, the casual group is probably closer to 5 than 150.
Delving into digital communities has proved to be a rich seam for researchers. While it might seem something of a trivial playground for anthropological research, online gaming communities give us the opportunity to view human behaviour aggregated at massive scale.
Research into Massively Multiplayer Online (MMO) games, such as World of Warcraft and Ultima Online have also suggested that the number of players actively involved in their player organisations (guilds) have a tendency towards the 50-60 player mark and maxing at around 150. Bolstering this evidence with data from social networks, research by Facebook’s resident sociologist, Dr Cameron Marlow found in 2009 that the average Facebook friend graph is 120 people.
Courtesy of Christopher Allen
Online games, as disparate as Warcraft and League of Legends, regularly have small player groups restricted to 5 or 6 players by design (just as basketball and soccer teams are limited in number). What’s striking is that, of course, these numbers very neatly echo Dunbar’s numbers as they diminish in size with increasing intimacy – guilds of 150, teams of 5.
Moving away from online games, and back to the real world, the author Malcolm Gladwell reported in his book Tipping Point that W. L. Gore and Associates (inventors of GORE-TEX and often considered one of the best companies in the world to work for), limits each of their physical office locations to 150 staff, having arrived at this magic number by ‘trial and error’.
From a personal perspective the Dunbar numbers are fascinating having worked in a company that grew from 5 staff, felt most unique and vibrant at 50 and rapidly passed Dunbar’s Number to reach around 350 today. Certainly, 150 people still felt like a manageable community, while 350 always provides an opportunity to make new acquaintances.
It’s not people, it’s relationships
But while all this talks about evidence that suggests that there are naturally comfortable numbers for our social groups, teams and companies, it hasn’t yet suggested why this might be the case. Perhaps the most compelling theory is that it’s not the number of people themselves that limit our ability to network, but rather the number of connections that explodes between all the people in that group.
The late J. Richard Hackman, a Professor of Social and Organizational Psychology at Harvard University, is best known for his research into team structure and performance. Not a man to mince his words, Hackman once bluntly stated, “Big teams usually wind up just wasting everybody’s time.”
Hackman proposed that the twin problems of motivation and communication led to problems in larger teams (or in fact, teams of any size). What Hackman found most important, though, is not the number of people but the number of links between them that accumulate when group size increases.
This growth is relatively easy to see when mapped as a graph (thanks Neo4j for being awesome)
The pentangle was unintentional, honest
In a small, 5 person group there are an impressive 20 relationships. Person 0 knows Persons 1 through 4, Person 1 knows 0, and 2 through 4. And so on, and so on.
Imagine this as a scenario at the pub – Person 0 buys the first round, and it’s all pretty manageable. By the end of everyone’s first round, 20 drinks have been bought. A pleasant evening was had by all, and most importantly, everyone in the group has a good chance of remembering whose round it is.
But if this group were to grow to 13 people – which, incidentally is the largest Scrum team we’ve ever temporarily allowed – the number of relationships grows to a not inconsiderable 78. Now run that pub scenario through your head again. The first round is 13 pints, and by the fourth person’s order there’s not a chance of remembering who still needs to take a turn at the bar. Issues of memory, co-ordination and even the physical proximity of the group are going to take over. You can almost guarantee that the group will split into smaller groups, that there’ll be an argument and someone will have a terrible night. That’s life.
It also turns out it’s maths – this growth in relationships is governed by a nice simple equation: r = (n^2-n) / 2. This broadly says that the number of relationships (r) will grow at a considerably faster rate than the number of people (n) – a power law relationship, which we can see in this graph.
Note – there’s a blue line along the bottom that barely moves off the scale as the relationships head off into orbit.
If we want to really demonstrate this, the graph below shows that the number of relationships generated by a notional team of 50 people is an enormous 1,225. Remember, this was a company size that I suggested had felt vibrant and still close-knit.
At 340 people, the graph of relationships grows big enough to crash my browser when drawing the visualisation – a staggering 57,630 relationships. At this point, well in excess of Dunbar’s Number, the pub scenario is a distant memory and the only option is to ensure that the larger group can be supported in splitting down into smaller social and working groups
Now we’re just getting silly
The TL:DR;
So, what’s the takeaway from this? When asked about how big a group should be, how should you answer?
Well, armed with nearly a century of empirical evidence, you can now feel relatively assured that smaller teams will outperform larger teams. You can also feel reasonably confident suggesting that the ‘magic’ number rests somewhere within the range of 4-9 people.
This range is a good rule of thumb, as it means that as the team is getting to the upper end of the range, it’s also the right size to split down into two groups. Fewer than 4 people, and the group may lose the skills, collaboration and creativity that make groups so powerful.
You can also feel reasonably confident suggesting that a group starts to reach about 150 – a departments or companies for instance – it might be time to start taking steps to address its size.
However as a parting but critical comment, I feel moved to remind you that, regardless of its size and all of this research, any team is only as good as the people that form it. Their drive, commitment, skills and suitability for the task at hand will always be bigger drivers than team size in predicting success.
Team size is just one, simple component of any team’s dynamic. So, let’s just try and keep the teams small and focus on nurturing the people.