More Thoughts on Web Analytics, Social Networking, and Social Networks….
I’ve been taking a look a deeper look at the trends in social networking and the analysis of social networks using nodes (such as taxonomy) and ties (such as clickstream data). A few concepts from networking theory are intriguing me, and I figured I’d bring them up here to see if anyone has any thoughts:
- Betweenness. Identifies the degree to which a node in a social network is interrelated to another node. Identifying degrees of betweenness in taxonomy nodes and combining with “normal” analytics data could enable the analyst to:
- Detect nodes with the most betweenness to identify content that should be *automatically* served when a visitor interacts with a related taxonomy node (extending site optimization technologies)
- Determine misappropriated editorial agenda and withering products by contrasting the “popularity” of nodes with the most or least betweenness.
- Clustering. A concept used to express how visits relate to core taxonomy nodes could:
- Provide a means for visualizing how visitor segments cluster around particular pages or nodes in a taxonomy
- Enable the analyst to visualize the broad content themes that drive the most visits
- Density. Certain bloggers and site pages tend to see larger numbers of repeat visitors, comments, or maximized time-based metrics when compared to other pages. Can a metric for “content density” of a site be calculated? Perhaps by crafting a equation from counting objects in a taxonomy node, value-scoring each object, and seeing which objects were interacted with most frequently?
- Influence. The guideline is 99% lurk and 1% influence. Can we gauge visit “influence” and visualize it from:
- Pathing where visitors who have performed the most/least interactions and contributions ”go next” off-site.
- Value scoring an “influence metric” for Interactions, Contributions, posts and comments, and off-site exit links in each visit, then adding up the values to calculate a new influence-based KPI measurement per visit. Finally comparing the “influence metric” across all visits.
If you are still following me ( 
), what I’m working at understanding and reconciling is whether social network analysis theory when combined with web analytics can illuminate the analyst with new ways for thinking about a web site.
By combining a rules-based approach to processing this type of data, the possibility for automatic content targeting and the idea of a “living site” self-optimizing based on visitor interactions with taxonomy nodes or site objects becomes closer to reality. The potential to use analytics data and social networking theory for building and realizing new combinations of product, content, and design becomes possible. For example, I could create rules and logic commanding my CMS fill a “related topics” module or widget on a particular page with content from nodes that have the smallest amount of betweenness and the greatest density.
It’s clear that social networking impacts web analytics. Most major analytics vendors don’t seem to be thinking about applying (or how to apply) concepts from social networking. I’m looking forward to vendors bringing social network theory into their technologies by perhaps combining, rules-based algorithms for site optimization with existing analytics data and new, open API’s (for example, Facebook’s new API or LinkedIn’s forthcoming API) to drive profitable revenue from new and existing channels.
Judah added the following ...
Hello Steve: Yes, I am aware that SNA software exists in the intelligence community and companies like Visual Sciences and BuzzLogic have the capability to visualize these types of relationships too. I’d like to see SNA concepts applied to web analytics objects to help the analyst uncover hidden or unknown relationships and to provide context for rules that could be developed to drive automatic site optimization. In this post I was trying to hint at how the mainstreaming of social networking, web analytics, site optimization technologies, advanced rules engines, and concepts like taxonomy/ontology could lead us to new techniques and products for “doing web analytics” and enabling internet strategy. Thanks for reading and your always informative and entertaining comments. Did I mention that I was the 146,xxx person to work for SUN?
CJ added the following ...
It would be nice to see this SNA stuff appear in commonly used web metrics packages, such as Omniture or HBX. If you can do pathing with pages, the idea of doing pathing, etc. using custom variables should be rather straightforward, with privacy regulations taken into account. I know you can already do correlations, for example, to see if a person of a particular job title (using cookies) is viewing particular types of web pages… this could be very useful for knowledge mgmt. purposes on intranets, or for managing ROI on extranets, as well as the marketing implications for community driven internet sites.
Marianina Chaplin added the following ...
Hi Judah, mainstreaming social networking and yes blogging and how they are impacting on our society, generation of ideas, groups, businesses are absolutely fascinating. I agree with you that Social network analysis theory can give us better ways to understand grouping and degrees of relativity between what people do online. I don’t know if you will agree with me but I think of web analytics as web anthropology - the context (ie the betweeness and clustering you mention) and relativity are so very key when looking at visitor observation and data. “Social anthropology has been distinguished from other social science disciplines by its emphasis on in-depth examination of context, cross-cultural comparisons (socio-cultural anthropology is by nature a comparative discipline), and the importance it places on long-term, experiential immersion in the area of research, often known as participant-observation. Cultural-Social anthropology in particular has emphasized cultural relativity and the use of their findings to frame cultural critiques.” By the way, I am a qualified anthropologist from University College London (Bsc). When I discovered web analytics, this will sound weird, but I felt giddy with excitement. The internet is the largest human interaction social networking experiment the globe has ever seen (more than cool). I am sure site content will be generated by betweenness and context (already happening to some extent with google personalised search for example) in the near future but it could become intrusive and big brotherish.
Judah added the following ...
CJ: Your point about “privacy” is absolutely correct especially when you are decoding cookie values to be things like job titles, people, or any type of PII (personally identifiable information). Using advanced web analytics tools you can generate reports on custom dimensions in your data - a powerful capability with many uses as you suggest. Good stuff.
Marianina: If content is king, then context is key. Relativity in web analytics is well represented by this conceptual notion of “uniqueness” don’t you think? I don’t think it’s weird at all that you get excited over the data and web analytics. In fact, I bet most web analysts really enjoy what they “do.” A friend of mine named Joseph Carrabis, who runs an unique company called NextStage Evolution (http://www.nextstagevolution.com/), has educational roots in anthropology. You may want to check out some of the work his company is doing if you are fascinated by such realms of internet analysis. Google seems to be doing very solid work in personalization using opt-in data from search queries. If consent and usage is well declared and opt-out procedures are easy, I don’t have a problem with companies personalizing or persona-lizing content or offers, but I am concerned with data security issues and sharing. It’s a fine line, and I’m curious to see how society will evolve this wonderful creation, this internet, throughout our lives.
Marianina Chaplin added the following ...
Hi Judah. Thanks and I’ll definitely check nextstage out.
Marianina Chaplin added the following ...
I also saw some really interesting work that MIT social media labs are doing measuring the true influence of myspace visitors with the true influence of comment flow. I’m definitely not saying this for linkbuilding/baiting purposes so if you are too busy, I understand but I just started my own blog Web Analytics Princess http://www.marianina.com and if you’re not too busy, I wrote a post on social networking analysis which I’d love to get your critique on. Don’t worry if you are too busy, I’ll still comment on your blog anyway.
Judah added the following ...
Thanks for the pointer to the recent work at http://www.media.mit.edu/ right next door in Cambridge, MA. I will read your post this weekend and comment. I have also added you to my blogroll. 
Steve added the following ...
Ha! I don’t know if you realise it Judah, but you’ve just re-described software that has been around for 15 years that I personally know of.
The software is typically used in the Intelligence/Policing communities. Trying to find linkages between disparate people or groups or the like based on limited information. It can be tenuous, but by being graphical an analyst can “see” connections that may otherwise not be obvious.
In your example picture, you can clearly see the links from Carl to Grandpa. And thus can see an alternate path for Homers illegal still to sell into the old-folks home. While Homer himself retains an immediate dissociation.
To knowingly mangle one of my favourite episodes.
In an added piece of irony for myself. That software ran on SunOS and eventually Solaris. My initial foray into the world of sysadmining unix servers.