Subscribe to Vincent Granville's Weekly Digest:

If you had to select one metric to measure the success of a social network, what would it be?

Difficult question, as it depends how success is defined. But as an advertiser, after identifying several targeted networks in my vertical, I would look at the following types of metrics:

  • average time to 10-th most recent visitor
  • average time to 50-th most recent visitor
  • average time to 10-th most recent NEW visitor

Also the advantages of these metrics is that they can be computed easily. An advertiser can run a little web crawler to automatically compute these metrics on target publishers (networks), rather than relying on 3rd party statistics that can be inaccurate. For instance, on AnalyticBridge, time to 5-th most recent NEW visitor is about 12 hours as of today. You can just look at the front page to manually compute the number (a web crawler would automate the computation and also compute averages over several days).

These are important metrics to get an idea of reach and frequency. Of course, would such metrics become popular, you would start seing some networks inflating they member counts by signing up fake members. But we are not there yet.

Views: 21

Replies to This Discussion

I have been thinking about this question in the context of the longevity of a network. For longevity you need to measure the participation rate of the members. I have been doing an unscientific sampling of different social networks ranging from political, scientific, and social. I have not found a network yet that after reading it for a week I make part of my daily routine. Either the quality of content is too low, or the frequency of updates is too low.

As an example let's look at AnalyticBridge. I tend to spent several hours on it every other week or so and typically I login in because I have a question. In researching the question I read lots of questions and replies and maybe contribute a comment here or there. However, after a couple of months my questions become more complicated and now AnalyticBridge becomes less useful to find the information or hints about venues to pursue to find it. The problem IMHO is that AnalyticBridge doesn't have sufficient content to make it part of my daily routine, and thus over time my focus will shift and I become a less productive member of the community.

I think the success of a social network has to be measured in terms of the prime goal of social networking and that is user contributed content. Given that this content can be popular or deep depending on the context of the network, the metric should capture that too. For example, content for a social networking site like Facebook could use a tournament/popularity contest metric, whereas a network like AnalyticBridge would need a something more substancial to measure quality of content.

What do you think the future holds for social networks? I am getting to the point where I no longer see much difference between a social network and the old newsgroup dynamics. For me the new car smell of social networks has worn off and I now realize that creating good content is hard. Creating consistently good content requires the same editorial structures that have developed at the newspaper and magazine industries.
It is difficult to produce good content, for a few reasons.

First, people don't want to post proprietary material. I've tried to attract companies to talk about themselves and the problems that they solve or they face, their success stories. I believe that there is some potential here. Also I believe that people will eventually talk more about their success stories.

Also, among the three members who were elected "Member of the Month" over the last three months, two have contributed a very good "Interview with Analytic Leaders". So, financially supporting the production of content has positive impacts. In the future, I might shift some of the advertising dollars into content production. I also wish I could contribute more personally in terms of answering questions, but like most of us: I'm very busy working on many things, and much of what I work on is proprietary. I do have a few interesting stuff in the pipeline that I will post over the next few weeks: reach and frequency mathematical formulas for online advertising, logic and PL regression, detection of bogus members in social networks, making social networks profitable, etc.

Some content is semi-automated (RSS feeds, job ads, conference announcements). I have increased the number of feeds -- I have added Techcrunch recently, as suggested by a member (see http://www.analyticbridge.com/group/news). At some point, I'll have to find a system for automated feed management and clustering.

More so than content, I believe that the value of a social network is networking, and so far I have spent more time in making AnalyticBridge a strong community for networking purposes rather than focusing on content production. But this might change as momentum increases. The last three days have been very good in terms of traffic and returning visitors, see the chart at http://www.analyticbridge.com/group/aboutanalyticbridge.

Finally, almost all posts are kept alive permanently. We have now more than 500 of them. When we reach 5000, the search box at the top of this page (right hand corner) will be very valuable to find interesting content. Also at that point, we'll have 5000 members, and be a very popular destination for analytic professionals. We should keep in mind it all started only 5 months ago. AnalyticBridge is still a "baby", but growing fast. Eventually several people will manage this network.
That is a good point, Linked-in is a great example of a successful social network that I don't go to for article content. I guess one could make the argument that linked-in's content is still user generated and it is their biographical information and it is the quality of that content that makes linked-in successful.

I like the idea of AnalyticBridge being a networking site with a stronger vertical focus. For that to work, we would need to capture more 'networking' data. And also analyze how people network in a vertical like analytics. I have been looking in AnalyticBridge for quantitative folks but have not found this to be productive (as of yet). So even in a directed vertical there still is a huge variability of people with varied tasts and needs. I would love to help you in thinking this through.

Personally, I feel Ning is very weak in the social network management respect. Ning feels more like a facebook wannabe than a Linked-in given the widgets they provide. We should look at some javascripting to provide more functionality.
I found LinkedIn's Q&A has great quality contents, rarely find anywhere else: high level of expert participation and low noise.
It depends. You also have people promoting their products, and people who think they are experts but know very little. I once asked a question (on LinkedIn) about how much revenue one can expect to make per thousand page views when you run your own advertising program and target a niche market. I've received 20 answers from 20 people, all wrong by a factor 1000. They were all assuming (despite what I wrote) that the way you generate ad revenue is by placing a google ad banner (running adsense). Not only can you do 10 times better by placing a google search box rather than an ad banner (and there's one on AnalyticBridge at the bottom of the front page), but if you actually run your own ad program, you can do better by several order of magnitudes. Nobody on LinkedIn would believe me, even though this was based on factual evidence from my own web sites and competitor websites.

One way to keep great quality content is by removing worthless discussions on a regular basis, and never deleting great discussions.
Exactly! Totally agree with the last paragraph: editoral control makes the difference. And what I find so interesting about that insight is that it is exactly the structure that developed in the old media companies, news papers and magazines. When they have a great editor, or editorial staff, these media outlets are fantastic.

In the world of media however we are seeing a worrisome development and that is the segmentation of the audience. I once heard it described as "putting a mirror in front of your target market" and creating cohesion through 'similarity opinion' instead of intellectual discourse. Due to the fact that is heaven for an advertiser to only pay for the target audience they seek, this is a business organization that is here to stay. Organizations such as News Corp are so large that they can still cover the whole spectrum through different focused properties. For example, they put on Bill O'Reilly who will be ranting about "The Simpsons" which is also an News Corp property.

To have a media property that can opine freely while covering the whole spectrum of humanity is going to be an interesting experiment.
How you can automatically filter our non valuable contributions is an interesting problem. I've thought at eliminating contributions after n days, where n varies between 0 and infinity, based on some criteria such as
  • contribution is 2-week old, less than 300 chars, and has no replies (n=15)
  • contribution is less than 30 chars, no links (n=2)
  • contribution is from same author, points to same web page (n=2)
  • job offer, non paid (n=30)
  • Contribution is just a link (n=?) (can be very valuable sometimes, but is more likely to be spam)
The best filtering will eventually be a combination of automated tools and human intervention. Another example of community, Craigslist, where filtering is performed by members, is not a success in my opinion, in terms of content quality.
In a deep knowledge site like AB good content is sometimes unpopular if measured in terms of links or comments. For example, your question on an iterative method for Linear Regression. I have read that half a dozen times, I start trying to add a comment that would add to that discussion but then something happens that distracts me again and nothing happens. Deep knowledge with good insight would be AB's most valuable content but it could not be measured in terms of Page Rank or other tournament style metrics. As an editor you have to recognize those and maybe the automation that would help here is a mechanism to isolate those posts and order them for human measurement.

Do you believe enough in SVM? Great content should match against great text books or seminal papers, or at least come near. If we could use an algorithm like SMO AND have a reference library to seed the classifiers this could be interesting.

In this context, I have to believe that NEC site seer has done some research on this. They have a lot of deep digital content and combined with traditional reference counting this could be a very valuable data source. Do you have any NEC site seer contacts? Are there any on AB?
This is not a reply per se. Just wanted to tell the two of you how interesting your conversation has been, especially regarding the evolution of social networks, news systems, etc. Transition is a constant and both of you seem so aware of that.

RSS

Follow us

© 2013   AnalyticBridge.com is a subsidiary and dedicated channel of Data Science Central LLC

Badges  |  Report an Issue  |  Terms of Service