- April
Posted By : Adrian Chan
Social capital on twitter: analytics of flow

I’ve been thinking lately about social analytics, and in particular how it applies to twitter. Twitter is a conversation tool, and content on twitter is more akin to speech than it is to its long-form brethren. The term “micro-blogging” is, I think, a bit of a misnomer in fact. “Micro-blogging” suggests writing (blogging). To me, twitter is clearly talk. Micro-messaging would be more accurate — but then messaging is micro already.

Because the content on twitter is produced by people talking, it would need to be measured differently than conventional page-based content, social or evergreen. We would want to measure talk, not pages. We would want to measure talkers, not sites or domains. We would want to measure relationships, not in-and out-bound links.

But I think the industry’s long legacy in web analytics and web traffic analysis will most likely result in early-generation tools built around the measurement of web traffic. That’s the easiest migration path for a web analytics to social analytics tool — repurpose existing methods and technologies. Visible Technologies, Radian6, Techrigy, and other tools tend to focus on traffic and enhance it with views of topical spaces, tag clouds, and volume around twitter. (See Jeremiah Owyang’s coverage of social media measurement.

Having used these tools, I could understand it if some of you have had the experience I have: there’s a lot of interesting stuff in there, some of which I wouldn’t have noticed, but I’m not sure what it means. The approach taken to blog and web site measurement, which hails from search engine approaches and is in fact closely tied to search engine results, maintains a focus on phrases and words. Value is then assigned to pages (content) by means of relative rankings. The relevance of a visitor or visitors (people) is imputed from clickpaths and search phrases. All of which paints a picture of people looking for something, which we assume is related to the words and phrases we have captured them using.

So to transfer a search-oriented methodology into a conversation space (twitter) seems misplaced and misguided. And may explain why at this point we have no idea what to make of twitter analytics other than to count people and posts relevant to us.

Measuring user activity in order to glean valuable information from it will fail if the measurement methodology is incommensurate with the activity taking place. If a tool’s tracking DNA was designed for a population of people looking for information, getting it from pages, and qualifying which ones are valuable by clicking or not, then the search at the core of that tool is misaligned to conversation spaces. In conversation spaces, or the social web, people talk to other people, expressing not searching, and addressing themselves to the public at large, to small groups (followers, peers, affinity groups, friends), or to individuals. The meaning of the words they use is not akin to the meaning of words found by search analytics; what’s measured will be misinterpreted by analysts if it is misunderstood as a query.

I believe we have to learn how to mine social capital and flows of social currency in conversation spaces. Users have an interest in gaining an audience. They want to accrue interest and get attention. They speak in ways that attract attention from strangers, and hold conversations with those they know. The fact that all of this occurs in an open social field creates a significant number of social distortions, yes, but those could be accounted for if tools were properly designed to filter out those distortions. An understanding of speech — in terms of statement types, addressing, response and uptake, distribution, and so on, would be much better suited to the space than the current and conventional analysis of word-based queries.

Relationships among speakers organize and inform how they talk and about what. Open social relationships have a different structure and organization than social networks. So this would have to be modeled and used as a means of making online talk relevant. The user centricity of the space would need to be accommodated, as talk in open social spaces has to do with establishing presence and soliciting the presence and attention of others — again, a kind of activity in sharp contrast to the use of words in search. Search phrases address the brand or information sought after, making the connection between user and results simple and direct. Words used by people to conduct open talk make an appeal to the attention of others — the direction of speech is in other words reversed: it doesn’t look for and query, but instead appeals and attracts.

Brands of course recognize that influence is involved, but continue to think of it as a property of a person, when in fact all influence (power) is a relation. It can be undone by an audience and in fact it “exists” only in the “eyes” of the beholder. Which is why influence needs to be maintained and sustained by talk and activity. It is not a property accumulated and protected, nor is it diminished when spent. Social capital in fact accrues to the person spending it, and its “expended” by communicative acts perpetuated and distributed by others. This has always been the case with dynamic social capital and status: it must be used if it is to be increased. And yet we continue to count influence by followers, numbers, and quantities that give us the false impression that it is a property owned, not an process sustained.

Time is the difference between the approach taken by existing metrics and those required for conversation spaces. Conversations and their dynamics are temporal. The social dynamics of conversation can only be understood as social proceedings that unfold in time, over time, and having temporal properties associated with human experiences of time: fast, slow, waiting, hastening, pausing… Pace, rhythm, and flow. I suspect that most currency traders would know what to do with this


  • Spot on:

    The current tools tally up mentions and then try to quantify the metrics, in the most misplaced and futile way. what good is this quantification and these false “tone” metrics? Does a Tweet conversation persist to a users blog entry, do the comments endure? Is there a growing momentum?

    None of the current tools addresses these things, and I am glad, Adrian, as one of the leading thinkers in REAL Social computing, that you have seen this for a long, long time.

  • Cheers Alan and great to hear from you! Coming from a wizard of analytics that's quite an accolade.

    If pandora is better at music analysis than anything else, then at this point we're just tweedling knobs on a band equalizer. Pandora's in the future for social media analytics. And after that, some strange social algorythmics built from collective social genius as well as smart song matching. You get my drift. Momentum, yes. Reach, yes. …

    I take some influence from Balinese water culture. 😉

  • I'm not sure where this space is going either. Right now we're just counting single words or phrases and accumulating that information over time, counting it, creating averages.

    Another real interesting area is charting how a meme/tweet moves across a social graph and is retweeted/broadcast. Showing how this is happening would say something about the quality and scope of memes and the social network as they move about.

    These are all just moving targets at this point, although we have considered mining this stuff at some point in the future.

  • I see social media as a new “means of production” and talk as a means of distribution. Mapping distribution and correlating to social graph for types of influence is where it's at (or will be at!)

Leave a Reply