For ages researchers have talked about averages. The average TV viewer, the average music buyer, the average newspaper reader and so on. This relies on the fact that the world follows a nice bell shaped curve where most people are clustered around the average. It’s an assumption that works when you’re talking about physical attributes and holds fairly well when talking about mass behaviour.
People in parts of the research industry have always been aware of the power law and power curves; more people are becoming aware of it all the time. It turns out that this describes much social interaction on the internet. It’s pretty simple to describe in this context – any nth person in a system is 1/nth as active as the most active user – but it screws with all researchers’ assumptions.
In anything described by a power law the most active person will be twice as active as the next most active person. The 100th most active person will be 1/100th as active as the most active person. What this means is that the mean level of activity for any person will be considerably higher than the median (the middle value) and far far higher than the mode (the most common).
So the most active tweeter may tweet 100 times a day. The average tweeter may tweet once a day but the most common tweeter may tweet once a month. How to describe a twitter use, then? Surely the most active cohort will be very different from someone who is right on the average – but the person who is right on the average is very different from someone who has the most common level of activity.
The current move to research communities and social listening tries to address this issue. It does a good job of describing maybe the first and second quartile by activity in any area – the people who are more active than the average (erm, quartile is a really bad way of looking at a power curve but it’s easier language). What I doubt it can really do is describe the third quartile – people more active than the most common but less active than the average. I’m sure it can’t really describe the most common user, people who not very active but who make up the majority of users by individuals, rather than by activity.
Does that matter? Maybe not if you’re talking about activity on the site itself but it sure as hell matters if you’re talking about opinion and behaviour in the world outside. If you extend this to another category, for example movies, then a community may tell you what the most active moviegoers are talking about but they won’t tell you what the person who goes twice a year is interested in. Does this matter? Not for 80% of the time but usually that 80% is what you could predict without too much hassle. It’s the 20% that you can’t predict and I’m not sure research communities in their current form can help you greatly.
Specialist communities need to be built within larger general communities. This could be where the larger research companies can fight back against smaller, lighter community specialists. Yes, you need to make a community up that will be driven largely by its most active users – but you also need to let people who are not part of your core community get involved when they wish.
Research companies could start seeing their general panels as a series of interconnected communities. If you need to, go out and recruit a specialist panel of movie-goers but embed it in your general research panel. Let people find it and contribute to it as they want. Why ring-fence them away from the rest of the world? That’s not how the world works.
Real communities involve everyone from the most active to the least active and the least active represent a much larger audience than the most active. The larger research companies need to learn from the specialist community builders in lots of ways but they need to recognise they have something these specialists don’t: They have the whole community. They can anchor specialist groups within the wider community to ensure that the huge breath of opinion out there is available to their clients, not just the most active.