Dunbar's Number, Span of Control and Lean Organization DesignBy Jon Miller | Post Date: June 21, 2011 9:21 PM | Comments: 3
A few weeks ago I learned about something called Dunbar's Number while listening to the radio. The relevance of Dunbar's number to lean organization design struck me immediately. There are such things as magic numbers. Some of these relate to statistics, such as sample sizes, or a minimum 30 repetitions when making observations of a process. Some of them we call them fundamental physical constants and most of the time these do their part in keeping things together unnoticed. Yet other numbers numbers deal with human cognition, behavior and social interaction. We are told that people can keep 7 things in memory, but struggle with more. Within organizations there is a rule of thumb for effective team size as being between 5 and 8 people, whether it be a natural work team or a project team. Dunbar's number takes this idea to the next level, possibly defining the maximum effective size of an organization. It needs to be said that this rests on the premise that stable social relationships are essential for effective organizations. These stable social relationships are based on social contact and mutual recognition. These in turn are limited by our cognitive capabilities. Dunbar's number hypothesizes that there is an upper limit beyond which detailed rules are necessary to maintain stable social order. Based on observations of villages, armies, academic organizations, communities and companies, this number is proposed to be around 150 people. In the words of British anthropologist Robin Dunbar "this limit is a direct function of relative neocortex size, and that this in turn limits group size ... the limit imposed by neocortical processing capacity is simply on the number of individuals with whom a stable inter-personal relationship can be maintained." Our brains are only capable of sticking together within a community of 150 or so, in other words. The example cited in the radio program was Gore-Tex, the manufacturer of wetsuits, hiking boots and ponchos. The founder Bill Gore set up the company in his back yard. The company grew and grew, until one day he walked into his factory "And he simply didn't know who everybody was." According to the radio program, Gore realized that after more than 150 people were placed in the same building, things at Gore-Tex did not run as smoothly because the sense of community was gone. Dunbar's number had been exceeded. As a result, Gore decided to limit his factories at 150 employees.
Business was never better. One-hundred fifty, it seemed, was a magic number. This theory begs some additional and rigorous testing before being placed into wide scale use, but the understanding of human cognition and social behavior is no doubt important to designing effective organizations. Applied to what we know from experience about lean organization design and specifically span of control of front line leaders, we can work out the following structure based on Dunbar's upper limit of 125:
5 group leaders each have 5 team leaders = 5 x 5 = 25 team leaders 25 team leaders each have x 5 team members = 5 x 25 = 125 team members grand total = 1 area manager + 5 group leaders + 25 team leaders + 125 team members = 156 total people
5 group leaders each have 4 team leaders = 5 x 5 = 20 team leaders 25 team leaders each have x 7 team members = 7 x 20 = 140 grand total = 1 area manager + 5 group leaders + 20 team leaders + 140 team members = 166 total people
One solution to that problem, he adds, can be seen in the modern military. Even as they create "supergroups" -- battalions, regiments, divisions -- most militaries are nonetheless able to maintain the sense of community felt at the 150-person company level. "The answer has to come out of that," Dunbar says, "trying to create a greater sense of community. In the end it's all about people, relationships, fairness and mutual respect. Formulas and models should not be applied too rigorously but instead used as guide lines for experimentation and learning. Humans may have tendencies based on our cognitive capabilities, but we must find ways to to move beyond our limitations, adapt and form greater communities based on a greater sense of common good. Dear Jon, grand total = 1 area manager + 5 group leaders + 20 team leaders + 175 team members = 201 total people Poster: John Santomer | Post Date: August 3, 2011 12:47 AM Thanks John I fixed the typo in the formula. Poster: Jon | Post Date: September 12, 2011 12:22 PM |




Jon, I could kiss you; I missed this NPR story.
I cannot tell you HOW LONG I've been looking for this! Going back as far as 6 years ago on my site, I've been talking about optimal size of apparel firms. [Apparel is all I know so I couldn't presume it applied to other organizations.] I've said a firm should not have more than 200 employees but that I thought a maximum of 100 was better. Lately I've been revising even that figure downward for lean domestic operations -again in apparel- to perhaps as few as 20.
I think humans do perceive a natural scale and order to things even if we struggle to articulate it (you do not know how long I've struggled to do just that). It cannot be a coincidence that 68% of women's apparel producers (associated with the need to rapidly change styles, much higher rate of new product launches annually etc) have 20 or fewer employees.
Through my own research, I've seen a dramatic decrease in profit per employee based on employee headcount the larger they get, especially in lean operations (I have one client who generates $250,000 in sales per employee!).
I have searched for analogies and an explanation for at least 15 years. In my struggle to articulate this concept, the only analogy I could find was related to Maslov's hierarchy of needs. I sensed there was a relationship there; that food, body covering and shelter industries were more related to each other than engineered goods. You can't pull coats and corn like cars meaning that some elements of batching is unavoidable in our industries and that even so, smaller entities were more efficient than large ones (ex. 80 acres are most efficient).
Thank you so much for writing about this, I can't thank you enough.