Blogging was where we began, and how we built our company so we have preserved this archive to show how our thinking developed over a decade of developing the use of social technology inside organisations

SNA tools: what are we measuring?

by

I have a pet project that I would like to share this year, and it involves analysing the evolution of a very active online community, so I have been looking round at SNA tools for the first time in a year or so. Social (and Organisational) Network Analysis (SNA or ONA) is a way of looking at a snapshot of communication data (usually just e-mails, but also documents and other artefacts) to gain insight into the connections between people and the concepts they are writing about. The idea is that this can tell us about ‘knowledge flows’ and the behaviour of ‘communities’ in large companies or networks. It is one of those things that sounds good in theory, but in practice
The defence and intelligence worlds are often seen as potential customers for SNA, for obvious (not very nice) reasons, but there is not really a body of evidence to suggest the intelligence community’s increasing reliance on SNA-derived insight has been successful. Indeed, it is possible that pursuing apparent ‘links’ based on computerised data and neglecting old-fashioned humint might be filling detention camps, but not necessarily improving awareness of real-world threats
In the commercial world, too, SNA can be an alluring idea. But if you are looking at your email store to provide knowledge or actionable insight into the present and the future, then I have some doubts. First of all, the data is based on a hypodimensional representation of existing practice, but not necessarily reality unless you assume that the communication and knowledge sharing tools used to produce this data were great at capturing it in the first place, which is not the case within most companies I know. Second, the nature of the inferences it makes can be very tenuous indeed. I suspect that if applied to my own email store, this approach would be completely unable to determine my most important relationships, because it is often those high-trust links that generate the least email traffic. Like many people, I have some very close friends and valued client contacts who send me less than a dozen emails a year, but I also have people who mail me several times a week trying to get a meeting. In addition, emails in the former category are often informal, jokey and sometimes written in a kind of personal code, whereas the latter are full of business buzzwords and seemingly professional prose. I have yet to see the SNA tool that can work out that conundrum, but I would like to use it as a spam filter if it exists
In the academic world, social network analysis is a multi-dimensional enquiry that tries to solve some difficult problems in areas as diverse as epidemiology and athropology, and does not usually rely on a single dataset to derive network relationships. As a general principle and a set of analyses, it is of undoubted value in working with online social networks, and I think we will need some good tools like Valdis Krebs’ InFlow software and other products to make sense of fast-growing networks in the future. Even Sharepoint boasts an SNA plugin
UK company Trampoline Systems have developed a new SNA product, and they recently released the Enron Explorer to demonstrate its features using the now-public Enron mail archives. It has a nice interface and is quite entertaining (in a voyeuristic sort of way); but aside from the funny bits, does it tell us much we didn’t already know? I cannot find any new insights into the Enron story uncovered by the tool. As Malcolm Gladwell argued recently, Enron was a puzzle not a mystery – it was just nobody had an interest in putting the relevant pieces together until some journalists started digging
SNA tools like this one also raise the question of whether they are more or less likely to make people share useful (or incriminating) information in the future. Several former ENRON employees have said the Trampoline tool adds to their own post-Enron humiliation, and presumably the majority of former employees are guilty of nothing more than ignorance. Trampoline’s defence – that they launched the site just for a few friends and were surprised by is popularity – is somewhat undermined by the invocation “Insider story from Enron? Tell the world!” at the top of the page. The dataset, which Trampoline are by no means the first to have used in this way, contains many non-Enron people who just happened to be copied into emails, such as recipients of press releases and spam or emails about fish. I am not sure about my views on the ethics of this, but I am clear about its likely impact, which will be to reduce rather than increase the level of meaningful interaction on corporate systems. If I worked in a company that might make decisions about me based on derived email analysis, then I would actively game the system through misdirection, which isn’t hard to do
Trampoline claim this is Enterprise 2.0, but I’m not so sure. Guesswork about social networks based on analysing email traffic sounds very much like one of the pipe dreams of old school Knowledge Management to me, and a long way from the emergent collaboration and SLATES that Andrew McAfee talked about in his original paper. Nonetheless, I am inclined to believe it is a really good, well-implemented SNA tool, and I hope it does well, but then I’m biased since they have an ex-headshifter on the team ;-
The issue here, I think, is that the best tools in the world are only worth the data that feed them. Trawling email and document stores in most large companies today would provide little more than a way of navigating by author, extracted keywords or inferred concepts, reflecting old structures and hierarchies that Enterprise 2.0 hopes to change. Useful in many situations, but not really actionable insight. I don’t think we can improve the data until we make it easier for people in companies to weave their own groups and networks and generate multiple layers of weak signals or ‘trails’ based on actual working behaviours. When blogs, social bookmarking, tagging and wiki-based collaboration groups reach critical mass inside large companies, then SNA tools will really have something much more meaningful to measure. Just today, thanks to Thomas Burg, I came across a good blog about the role of tag clouds in developing navigation based on shared meaning, and it reinforced my belief that networking based on what people choose to share is more efficient than broad-spectrum email text mining
We should have a deeper look at some SNA tools and think about how to apply them to the trails left by some of the online social networks we are developing. Maybe we have some use cases for that. Juan? Matt V?

5 Responses to SNA tools: what are we measuring?

  1. By ana on January 23, 2007 at 9:08 pm

    This is a very valid point and something which also intrigues me. How can you use technologu to sift through the trails of our online life (both work and social) to extract the links, the knowledge, the information, that can then help us become more efficient as individuals and help organisations to capitalise on their people.
    However, and although technology and adequate processes will come (no doubt) to help us do that, we can do SNA in the old, traditional, non-technical way. It’s not as quick, not as time consuming, but possibly more effective. Questionnaires with follow up interviews can work a treat to unveil hidden networks, patterns, obstacles, opportunities, gaps, etc.. You draw the networks and you understand them. It has drawbacks, such as the limit of numbers you should include, the way you can (or cannot) expect participation from all the people who would make sense to include (especially if they are out of the boundaries of the network you want to analyse or if you don’t know who they are).
    Via technology or not, SNA is great and a really good instrument to support knowledge management and social engagement activity within organisations.

  2. By Lars Plougmann on January 25, 2007 at 11:23 am

    ‘hypodimensional’ – I love it. Only nine hits on Google for hypodimensional, including this post.

  3. By Hans Henrik H. Heming on January 26, 2007 at 8:30 pm

    Hi Lee – interesting post. Thank you. Do you have any recommendations on tools

  4. By Charles Armstrong on January 31, 2007 at 4:56 pm

    Hallo Lee – Thanks for the coverage! However may I gently point out that the Enron Explorer site you wrote about is a technology testbed, not an enterprise product (as you seem to believe), so some of your conclusions are inevitably wonky.
    Trampoline’s main enterprise product is actually the SONAR platform. It sounds like you hadn’t seen this technology prior to writing your article. There are some details at http://www.trampolinesystems.com/products/sonar and if you’re interested I’d be delighted to give you a demo. We’d really welcome any comments you have once you’ve seen it.

  5. By Lee on February 2, 2007 at 8:27 pm

    Hans: Valis Krebs tool is worth googling, but in terms of mapping tools there are more and more open source tools around that will do this. The question is: against what data? Trampoline’s system presumably uses some form of clever text analysis and concept extraction against emails, docs, etc. to derive SNA information, but it probably has a few tricks up its sleeve as well.
    Charles: I thought I had mentioned that Enron Explorer is a demo of some capabilities of SONAR. If that’s not the case then please feel free to correct me. As I mentioned at the end of the article, I would like to have a look and, if it is good, perhaps use it.
    My interest is how auto-extracted social network insights compare with the trails and signals left by social media (and they are not mutually exclusive, I guess). If SONAR does this, I’ll take two 😉