FEATURE1 February 2010

The size of the matter

‘Dunbar’s Number’ is a theory that the human brain can only manage a maximum of 150 friendships. Robin Dunbar, who posited the theory, is currently updating his work to see if the Facebook effect has made any difference. Here – in an article first published in 2007 – Jamie Hamilton applies Dunbar’s learnings to online research communities.

What is the optimum size for a successful research-oriented online community? In the boardroom most would feel a little embarrassed proposing anything less than 500, and many would consider upwards of 10,000 perfectly reasonable. Yet on what assumptions are such numbers based, and more importantly, are they reliable?

Empirical evidence derived from neurology, anthropology, evolutionary biology and multiplayer online games would suggest not. Indeed, the findings of University College Professor Robin Dunbar and social software technologist Christopher Allen show that pitching any higher than 300 members will either result in failure or in something quite different from what was originally intended.

Dunbar proposes that the size of our social networks is determined by our brain capacity, which limits the number of people with whom we can maintain stable relationships. Using field studies of 36 primates, he discovered a correlation between the size of each species’ neocortex and the number in its social group. When he applied his formula to Homo Sapiens, it predicted 147.8 as a mean group size. ‘Dunbar’s Number’ has since found considerable support from census data on villages, tribes, academic communities, the armed forces and personal social networks across many cultures.

Shrinking expectations
Our experience of our own social networks – if we’re honest – is all it takes to corroborate Dunbar’s claims. So why does a membership of 147.8 suddenly seem pointless and lacking in ambition when it comes to planning an online community? For one, the vastness of the internet universe (and to a lesser extent, online panels) has a tendency to play tricks on the mind, rendering small numbers unthinkable and enormous numbers manageable. Furthermore, ‘community’ is an ill-defined term, used too freely. That it is applied to MySpace, Bebo, Facebook and other such entities is probably the root of the association between online community size and success. Yet these phenomena are not single goal-directed entities, but sprawling tangles of individual networks without clearly defined boundaries, identities or collective goals.

Scale may be seductive, but to be of genuine practical use to researchers and their clients an online community needs to be focused, purposeful and manageable. Dunbar shows us that this means we need to start thinking much smaller. In actual fact, he deemed 147.8 to be at the top end of the spectrum of social group size, a number associated with survival-oriented groups such as farming communities or military troops. Without physical closeness and frequent conversation (‘social grooming’) he argued, we should expect a mean that was lower still. Paradoxically, two of the principle attractions of online communities to researchers and participants are that they have no geographical limitations, and allow for irregular commitment, yet according to Dunbar these are not factors that favour the formation of large social groups. So how large a community can online conditions support, and are there examples from which researchers can take a lead?

Serious fun
The answers are to be found in an internet realm with much to teach online community builders: MMOGs or ‘massively multiplayer online games’. MMOGs are computer games with huge international memberships, capable of supporting thousands of users simultaneously. Perhaps the best known, World of Warcraft has over 9 million players worldwide. MMOGs involve immense fantasy worlds, where player-created characters are able to interact meaningfully with each other on a grand scale. Much of the gameplay operates at the level of a team or guild: tight-knit groups of players who get together on a regular basis to undertake quests and tasks. Such groups have many of the hallmarks of genuine communities: they are highly social, goal-oriented, and often have remarkably active, committed and stable memberships.

In his blog “Life With Alacrity”, entrepreneur and social software technologist, Christopher Allen, has compiled some statistics about group sizes in MMOGs, which he discusses in relation to Dunbar. As might be expected, several studies do reveal group sizes tailing off at 150, but more interesting is the finding that the vast majority are between 25-80 strong, with a mean of around 50 members. Allen believes these figures define the optimal size for an active “creative or technical group”, regardless of whether the context is online or offline. Anything higher than 80, he argues, requires so much socialisation to maintain group cohesion that members are left with insufficient time to devote to whatever reason they got together in the first place. Anything lower than 25 and there isn’t enough variety in the group to achieve critical mass.

Nevertheless, there is an exception to this rule that merits a mention – The Eternal City, an MMOG with a group size that regularly doubles up on Dunbar’s 147.8. Allen’s explanation is that the game is less social and more achievement-oriented than most, with interaction concentrated more on the environment than other players. This also suggests that there may be circumstances where researchers might justifiably violate Dunbar’s limit, for example, where their online community was heavily activity-based, or focused on a clear-set goal, but this is conjecture and requires further investigation.

All in all, however, there is little doubt that a research-oriented online community needs to be a great deal smaller than is generally assumed. The lessons provided by Dunbar and MMOGs are that to have a successful, stable group, we should aim at a membership of around 80, and never any higher than 300. These are limits set not by cowardice or lack of imagination, but by human physiology and behaviour. Many will ask if something so small is worth the effort. However, we should be careful that the glister of the grand schemes doesn’t blind us to the real benefits and unique opportunities of having a group of 80 well-screened, active, geographically-dispersed participants, immediately to hand, for an extended period. From a face-to-face qual perspective, this is a sizeable and diverse sample, which to engage repeatedly on an ongoing basis would be prohibitively expensive. Such communities may not be as mighty as we’d dreamed, but they work, they don’t require significant investment, and they offer researchers and their clients a set of tools, methods and possibilities they’ve never had before.

Jamie Hamilton is managing director of Nqual, an online qualitative software solutions and services provider. www.netfluential.com 

Editor’s note: This article was originally published in August 2007 and has been republished in light of recent interest in how Dunbar’s theory stands up in a massively socially-networked world.


14 years ago

Jamie, I think you've made some excellent points here, and I'd like to second and support them with some data. We at Communispace have run over 325 market research online communities of typically 300-400 members, and have amassed a body of evidence demonstrating that “going small” delivers large value for companies. First, though, it is important to arrive at a shared definition of community. We think of it as any group of people who share common interests, interact with one another and form ongoing, reciprocal relationships over time. While this definition is broad enough to include many social networks, it emphasizes the fundamental qualifier of relationship, which is critical when a company’s goal is insight, loyalty or advocacy. Participation and engagement in market research communities are directly tied to members feeling known and heard, as well as to a sense of exclusivity that comes from being part of a small, intimate group. Research we conducted a few years ago demonstrates that participation rates in small, private and branded communities outperform larger, panel-based or public communities in terms of average monthly participation, volume of posts, and weekly lurker rates. For example, in our communities only 14 percent of members on average logged in and did not post new content every week (or “lurked”). In contrast, this ratio is reversed for large sites where only one percent of site visitors create new content, another ten percent rate or edit that content, and 89 percent of visitors opt to remain silent and passively read. These disparities underscore the value of private, intimate settings for eliciting conversation among the majority of participants and hearing their unique voices. While public sites may garner more eyeballs, the vast majority of visitors to these sites are modestly engaged in the conversation at best. We've got quality metrics as well that indicate that the word counts and candor of postings in smaller communities also exceeds what's typical on sites with thousands of visitors. Ultimately, deep and actionable insight is contingent on trust, i.e. on community members trusting their fellow participants enough to reveal themselves, and trusting that you -- the researcher -- will be an active, transparent, and reliable listener. So thanks again for your engaging and eloquent article. And if you'd like to see more of the research I cited above, you can find it at http://www.communispace.com/research/about-communities/.

Like Report

14 years ago

This is interesting. However, I didn't understand most of it. To be honest, I am a badger and a community sett of about 50 gives me an idea. We've been around for a lot longer than you too so I guess we should know. Good luck.

Like Report

14 years ago

A really interesting read Jamie. I think one very important distinction here is what we mean when we talk about members. For me, the number of people registered to the community is in many ways an irrelevant figure. The key number is how many are active participants within the Community. And when talking about that group I think you and Julie are absolutely right…a smaller and engaged group are clearly the most productive. We’ve been running insight communities in the UK for a number of years now and are finding that the 300 active gives us richness and depth of insight on each brief we share. It also means that we are not reliant on every participant contributing to every topic – and this is another key benefit of going to the further reaches of the number of actives. Rather, participants can contribute at their convenience to topics of interest or relevance to them. It also means there’s an ebb & flow to the contributing group. So while the n= is important to us because we then know from experience how many will be active within that group, it is not a figure that we should focus on when talking about the size of the Community. Far more important, is what we do when we’re in there. I blogged about my concerns in July last year that we’re in danger of defining the community vs panel debate by size alone (http://www.virtualsurveysdiscussion.com/vslblog/?p=125). Another point I’d make is that some of this is very much down to what the client requirement is. So if the requirement is for a platform that allows both survey and discussion, size does come in to play. This Community-Panel approach has had notable successes that have been reported at conferences and in papers, such as that for A&N Media presented at last year’s online research conference. In these instances, again your principle holds firm with smaller more engaged groups being the most productive, but these can be ring-fenced or drawn out of a Community-Panel for particular tasks. They can also contribute to the whole. Julie’s comments about engagement are great too. We often find that social interaction actually isn’t something our participants necessarily want from a community of this type (we know, because we regularly talk to them about it). It is a nice secondary effect but they often socialise elsewhere online. Far more important is that they can interact with the brand. The community is an enabler to a conversation. And this is one of the key differences between a community for insight purposes and one that occurs naturally. The social glue of a community is to an extent imposed (often customer of a certain brand) and participants are then to effect change for the better. In many ways this is far more powerful a motive than a naturally occurring community. Sure, we make it fun and engaging along the way (that’s a hygiene factor of what we do), but ultimately we only succeed if we ensure a clear line of two-way communication between brand and participant. It’s this that means our incentives can be a small token of thanks in the form of a prize draw rather than paying by the response. This is a key point because Dunbar’s Number is about the human capacity to cope with networks…and in this circumstance assumes that you are interacting with all other members of the Community.

Like Report