Comment

Comment by Theodore Omtzigt on February 18, 2010 at 7:51pm

Terri Rylander blogged on Smart Data collective about Socialytics.

Socialytics

What in particular caught my eye was David Bakken's three dimensions of Socialytics

"Content: Just what is it that is being talked about. Could be brand or product mentions, could be sentiment. What are people saying?

Source: Who is generating the content? How does the content vary by characteristics of the source. He mentions the difficulty of knowing more about the “who” and suggests companies like Facebook might sell that information as a new source of revenue.

Connectivity: Who is talking to whom? This is probably most about influence and the value of networks — something that has proven very difficult to measure prior to social networking."

The piece we were talking about, how valuable is stuff, is an additional layer on top of these base analytics.

Comment by Theodore Omtzigt on February 16, 2010 at 12:13pm

I did that same connection check, and I am several one-hops away from the president of the USA. However, that still does not give me any access to the man, or to influence his inner circle. Even with folks that I am directly connected with, there still is a HUGE difference in how well I can influence these people.

Bottom-line is that first order statistics from social networks are just the beginning to characterize the 'value' of those connections. Particularly, social networks like this one, that are based on shared professional interests, analysis needs to dig a lot deeper, to quantify link quality. For example, all of us are linked to you, Vincent Granville. What is interesting to me is that the first order social network statistic will state that you as a network facilitator and aggregator are the most valuable 'node' in the analyticbridge network. From our (user) perspective our link to you is more valuable than your link to each one of us individually. Unless, there is a business deal you can get out of that link. That information is not easily captured by this social network: it can only be discovered through information collation and some stochastic analysis of deal flow/financial interrelation (IMHO).

Algorithms like PageRank do not capture that value very well: they only capture the popularity element and quickly deteriorate when big social/aggregator nodes enter the chain. I have been thinking and experimenting with different models of link value, but it seems to quickly deteriorate in an amalgamation of different models to try to capture different opportunities.

Is anybody aware of research in this context? The NetFlix price winning algorithm also deteriorated in a checker board of different models to capture that elusive next level of model performance.