Feb 17, 2010

Spread of Influence in a Network

About a week ago I put up a simple quiz on TwitPic using David Krackhardt's kite network as focal point. The kite network above shows a small group of people with strong symmetric[two-way] communication links. I asked,
"Where would you plant your msg in this net? Why?"

Several of my Twitter followers immediately answered and then the post was re-tweeted by several friends of mine. One friend sent it out to 30+ "social media mavens" -- none of them braved an answer.

This is a toy problem, yet it helps us think about how information and influence spreads in a human network. When I present this problem during one of my many talks on this subject, the first answer from many voices in the audience is usually "Diane." Then there is a period of silence and a few people sheepishly offer "Heather." Finally some joker in the back of the room yells out "Jane" and everyone has a good laugh.

So, what is the right answer? There are several.

The most popular answer of "Diane" is not a bad answer. The eye is attracted to the hub structure around her. She has the most connections and does reach a majority of the network with a direct connection.

The choice of "Heather" is a good one -- my preference. While Heather has only three direct ties, she reaches everyone in the network within two steps. Diane has several longer paths to reach everyone. Information and influence both degrade with each step in a network. After one step the message begins to grow fuzzy, after two it is becoming very noisy, and after three it is basically useless -- background hum. We might be all separated by six degrees but it is the first two steps that really matter.

Another good answer is "Fernando or Garth". They are between Diane and Heather and can also reach many people in the network quickly. Those that know social network analysis come up with this answer because these two guys have the best closeness centrality.

All of the good choices mentioned above are not guaranteed to get your message passed around even this small network. Just like a forest fire depends on one burning tree igniting another tree, or two, Diane, Heather, Fernando, and Garth all depend upon others to continue passing the message. It is not just the seeded node that matters, but the network neighborhood that the seed is embedded in! And... each node has a different threshold of adoption -- for one topic Carol may be a slow adopter, while Ed may be quick, and vice versa for a different topic/idea.

I am reminded of this song by the Alan Parsons Project -- The Turn of a Friendly Card:
"The game never ends, when your whole world depends,
on the turn of a friendly card"

Or in today's world -- the turn of a friendly tweet!

We have a simple problem, with no simple answer. So, how do you work this?

What happens when we try to scale this to real human networks that have dozens or hundreds of interconnected friends or colleagues in a network like below?

The secret is... redundancy! Yes, redundancy, that concept that we tried to eliminate in the 1990s with untold hours and dollars of business process re-engineering. Some redundancy actually helps networks function better.

In the simple kite network above we would use redundancy to seed the message with Diane and Heather! Some people may not hear Diane today, but will pay attention to Heather tomorrow.

In the real human network above we might need to find a dozen or more places to plant our viral visitor. Social network analysis software can help us discover the best soil for planting!

So, I tweeted the link to this blog post when I finished writing it: evening Eastern Standard Time in USA [GMT -5]. The response was not great. bit.ly showed me that 25 people clicked on the link in the first hour. The next morning I tweeted the link again and now more of my Twitter followers were paying attention -- by Noon bit.ly reported over 300 clicks. Finally I ran another tweet @ midnight, for all of Pacific folks, and early-risers in EU and RU. This added another 100 clicks accroding to bit.ly.

Lesson learned: It is not only the persons you plant the message with, but also the timing of the message. People pay attention at different times -- especially on Twitter. Redundancy in timing works too.


  1. Terrific, Valdis!

    I enjoyed your social network quiz, reading your explanation and learning how these connections work in a much larger network. My answer was Fernando or Garth. It was interesting to see that there were several good answers. I was happy to take a stab at your "fun" exercise!

    Thank you; it is always a pleasure to read and learn from your thoughts and analysis,

  2. An excellent exercise, and one that has generated much offline chatter here in Prince Edward Island. Thank-you for sharing the "answer" with us as I wasn't sure if we were going know for sure. Your teaser exercise couldn't have been better timed!

    Keep up the thrilling and fascinating work on social network analysis.

  3. It's always been amusing to me to see folks frowning on retweets, hashtags and syndication of content across multiple platforms. Yet those same folks promote Social Media for business as a marketing tool. Some even offer their advice on a consultative basis.

    For me, each of the aforementioned functions, or actions, makes one's "soil" fertile creating a productive, valuable node on the network. Unless, of course, they are syndicating only their own creations and information.

    It just seems natural for me to share other's creations. From my perspective, it's a win-win situation. "I" receive incremental attention while "you" and your content receives a gravitation slingshot. Works for me.

  4. You talk about adding timing for different time zones on Twitter. In real life don't we have to take another level of complexity into account? For example, a bit of text, an attitude or obesity might all travel along a network but use different channels: Twitter, phone, going for walks together, or eating. I wonder how these layers might be represented in a social network diagram? Does anybody try to do that?

  5. Welcome to the world of media planners ;-)

  6. Yes, we can map different links for different media, or different relationship types. The two diagrams above show work relationships amongst colleagues using any media.

    The details mapped depend on the project and the client needs. We can filter or slice & dice an organization or community in many ways!

  7. Thanks Valdis for re-establishing the value of redundancy in networks. It is an incredibly important asset.

  8. Valdis,
    This reminds me of the marketing phrase "reach x frequency". The efficiency is that you reach a targeted audience instead of the entire market, and then you let the network reach the rest of that market.

  9. Interesting post, thank you. It inspired me to write my further thoughts on this idea: http://bit.ly/byefCz