Sunday, July 12, 2015

Coolfarming Ideas through Knowledge Flow Optimization - Boosting Organizational Performance through E-Mail Social Network Analysis

Over the last fifteen years our research group at the MIT Center for Collective Intelligence, University of Cologne and University of Applied Sciences Northwestern Switzerland (FHNW) has studied hundreds of organizations through the lens of their social networks, extracted from the organization’s e-mail archive.

Among many others we have studied R&D organizations at car manufacturers, marketing departments at banks, sales teams at high tech manufacturers, medical researchers and doctors at large hospitals, and service delivery teams at large consulting and service provider firms. In addition, we have also looked at collaboration in open source organizations like Eclipse software developers, Wikipedians, and online communities on Facebook and elsewhere.

We have developed a 4-Step Process which we call “Knowledge Flow Optimization” to study and increase the performance of organizations, to “coolfarm ideas” (see figure below).

It consists of the four steps “Analyze – Predict – Mirror – Optimize”. To illustrate our approach, I describe the analysis of a fortune 500 high-tech company, where we compared e-mail communication of the organization with sales success of their sales teams in the different geographical regions.

Step 1: Determining Social Network Metrics and Communication Patterns
In the first step we analyze and quantify the communication patterns and social network structure embedded within organizational communication archives such as email, video conferencing and instant messaging. Quantified communication patterns include metrics such as average response time to messages, sentiment and contribution index. Contribution index is a measure of the balance of communication in terms of the messages sent and received by an individual. These are complemented by metrics computed using Social Network Analysis that are measures of social influence (or centrality) and their trends over time.

Step 2: Honest Signals: Comparing structural attributes with business success
In the second step we compare communication behavior found in step 1 with communication patterns that we have identified over a period of 12 years in over a hundred ONA projects carried out by our team. These patterns, also called “honest signals”, are indicators of better connectivity, interactivity and sharing among the individuals in the network. There are 6 honest signals that we look for, namely: Central Leaders, Rotating Leadership, Balanced Contribution, Rapid Response, Honest Sentiment and Innovative Language. Having calculated these honest signals from the data in the communication archives, we then correlate them with quantified success and failure criteria. The success and failure criteria vary significantly depending on the type of organization, the industry and the individuals being measured. In this example we measured sales performance of the sales teams in different geographic regions and for different products.

Step 3: Virtual Mirroring
In the next step we mirror the communication behavior we have identified for the different parts of the organization back to the teams and individuals. By showing them how they differ from the best practices we found in past projects, we help them to improve their behavior for better performance. Just like with a real mirror, looking at how a team “really” communicates can be an eye-opening experience for the team members, leading to fundamental changes in their behavior for the better.

Step 4: Devising a plan to optimize communication for greater success
Once we figure out which of the honest signals are correlated with success and failure, we developed a roadmap for the company to change sales communication behaviors leading to more successfully closed deals and more satisfied customers.

Why it is different?
When we do our analysis, we frequently get asked about the details of our method. There are a few key principles that I would like to point out that set our approach apart from other e-mail based and SNA approaches

Measuring “True Creativity” – our framework is based on the notion of COINs (Collaborative Innovation Networks). It has been field-tested in over hundred organizations to identify the communication patterns indicative of creativity. This includes far more than simply counting the number of e-mails of individuals and teams, rather, using the six honest signals of creativity listed above we have identified complex networking patterns of true creativity.

Know Cool People, and not just Hotspots – our semantic social network analysis tool Condor finds trends by finding the trendsetters. In the first step it works like Google, to identify who is using novel words and ideas, but then it finds the cool people, by measuring who uses novel words first, and how quickly they are picked up by others to grow into new COINs.

Anti gaming  - We are using social network analysis metrics such as “betweenness centrality” and time series of e-mail exchange, which are far more robust towards “gaming” by employees than simply counting e-mails sent and received.

Measuring organizational trust and satisfaction - We are not just counting complex words, to measure complexity in dialogue, and counting positive and negative words such as “great”, “wonderful”, “horrible”, “awful”, but through machine learning algorithms track word distribution and model positivity and negativity in context.

Understanding communication galaxies – we track the evolution of network positions of people, measuring how individuals change from being “stars” to becoming “galaxies”, as the most creative people and most highly functioning teams act as communication galaxies embedded into clusters of other teams.

Using E-Mail based Social Network Analysis gives an organization an unprecedented view into the nervous system of the organization, and allows it to predict flash points before they happen, leading to greatly improved performance and reduction in risk.

Sunday, June 07, 2015

Swarms do “what is right”

In a public vote at the town hall meeting on May 4, 2015, the Swiss town of Duernten (population 7000) decided to return $250,000 to a worker, who had not filled out his tax declaration since 1995 due to dyslexia. In successively higher tax bills the tax authorities had overcharged Ernst Suter by $280,000, forcing him to sell land to pay taxes on money he had never made. As a worker in a butcher’s shop Suter made about 60,000 a year, which, by Swiss standards, is a fairly low salary. Based on information from the town administration, the tax office of the canton of Zurich had second guessed Suter’s income at about 300,000 per year, and sent him tax bills accordingly, which Suter always paid, nearly forcing him into bankruptcy more than once. Only when Suter ran totally dry, did the branch of the town government that has the task to collect late payments (called “Betreibungsamt”, office of payment enforcement in Switzerland) pass on the case to a custodian, who brought the whole tragedy to light.

This is where things took a positive turn for Ernst Suter. After the custodian sent a detailed list of the overpayments to all citizens in Duernten, the town hall meeting decided last December to return the money. When the mayor of the town and his tax administration dragged their feet and decided to cut the return into half – because in the Swiss system, half of the collected town tax of individuals is passed on to the canton – citizens of Duernten took the matter into their own hands. They again put the issue on the agenda of their town hall meeting in June, voting overwhelmingly for Suter.

Fairness and ethical behavior is more important than strict adherence to the letter of the law. In this case, the swarm – the town hall meeting - also censored its leaders, asking the elected town officials to not only return the 250,000 they had taken from Suter, but even to make sure that in case the canton would want to tax the return payment as income, to allocate an additional 75,000 to pay for the income tax. They also asked the mayor to formally apologize to Suter.

This shows that the swarm (citizens of Duernten) knows better than the experts (the tax authorities) what is fair, independent of what the tax code says.

This case lead to intensive echo in the Swiss press (in German)
original exposure 
Series of articles in "Blick"

Thursday, April 09, 2015

Galaxy-Scope: Finding your virtual tribe (for example near the PayPal Mafia)

Whether it’s sitting in the same restaurant as George Clooney, or being on a picture with Warren Buffet, we are defined by whom we know and derive great satisfaction by being close to celebrities. Thanks to the Web and social media the six degrees of separation that separate any two people are shrinking rapidly. In addition, we can use the same insights to define who we are by looking at whom we are close to.

The following example illustrates how this idea can be applied to measure how close any aspiring Internet entrepreneur is to the “PayPal Mafia”.  In an article on April 1, 2015, the NYT describes the far reaching influence of PayPal alums in Silicon Valley. I was curious to measure the influence of the names listed in the article: 
Chad Hurley 
David Sacks 
Elon Musk 
Jawed Karim 
Jeremy Stoppelman 
Keith Rabois 
Max Levchin 
Peter Thiel 
Reid Hoffman 
Roelof Botha 
Russel Simmons 
Scott Banister 
Steve Chen 

I plugged them into our network data collectors for Wikipedia, the Web, and on Twitter. I did a Condor Coolhunting on the three infospheres using the names of the people as search terms.

The picture below shows their Wikipedia network:

Peter Thiel and Elon Musk lead the group, the rest is clearly recognizable, but none of them stands out. Thiel is close to Facebook, Musk to SpaceX. Chen and Hurley are close to YouTube, Levchin close to Yahoo, Stoppelmann and Simmons to Yelp.

The next pictures shows their importance in the Web:

The Degree-of-separation search uses the Google CSE API to collect the top 20 search results for each member of the PayPal Mafia, and then the top 20 links pointing back to each of the search results. Measuring the betweenness of the search results in the resulting bi-modal graph gives a proxy for the importance of each PayPal Mafia member in the Blogosphere as well as the most important Websites. Reid Hoffmann, Chard Hurley, Peter Thiel and Elon Musk are all similarly central, while and are the most central sites.

The Web Co-occurrence network shown above is constructed from the 3134 pages collected with the degree-of-separation search described above. Using named entity recognition and natural language processing, all people names are extracted from the thousands of pages collected. A link between two names is drawn if the two people are on the same page  - literally speaking. In this network, people like Barack Obama, Hillary Clinton, Steve Jobs, and Jon Stewart are more central than the members of the PayPal Mafia who were used to construct the network.

For the Twitter network, we combined the Tweets made of the members of the paypal mafia with all the tweets about them. An actor is a person tweeting, a link between two actors is drawn if a tweet is retweeted. Some people are very active tweeters, but are not so much tweeted about, others, like Peter Thiel, only tweeted once, but still has 90,000 followers, and is much tweeted about. And then there is Elon Musk, who does not tweet that much, but increased the value of his company Tesla by one billion with a single tweet.  By combining the two input sources, we get a Twitter network reflecting the real importance of the PayPal Mafia in the Twittersphere. It turns out that Peter Thiel and Elon Musk again rule the roost.

In the final picture we combined all of these networks (Wikipedia, Weblinks, Web Co-occurrence, Twitter). Peter Thiel and Elon Musk are the most important, taking their betweenness centrality as a proxy of importance. Compared to these two, all other members have considerably lower betweenness centrality.

As a second step, I was curious to see how ordinary people would fit in. Knowing fully well that nobody is “ordinary”, and everybody is “special”, this should tell us the “specialty” of each person in the context of big data and social networking entrepreneurship, also giving a metric on how important they are, and how close they are to the luminaries of the paypal mafia.

Eating my own dogfood, I started with myself. The picture below shows my personal network, cooked by the same recipe, combining my Wikipedia, Weblinks, Web Co-occurrence, and Twitter networks

As I do not have a Wikipedia entry, the pages on “Collaborative Innovation Networks” and “Coolhunting”, where I am mentioned, are the most central in the Wikpedia network. Also, as a passive tweeter, my tweet network is very small, so it is mostly the Web and Web content network that define my presence. That Barack Obama shows up, does not really mean that I have a personal relationship (I have not), but that we show up in the same texts occasionally.
The next picture combines my network with the network of the Paypal mafia.

I am not very close to any luminary, the happiness magazine is a surprising link (I do some research on human happiness, but was not aware of the link). Zooming in by eliminating the nodes with non-normalized betweenness lower 100,000 leads to the following network.

This is now much clearer, illustrating that my main presence on the Internet is the Collaborative Innovation Network entry, google Scholar, and arXiv, among others, plus a few common links with tweeters, linkedIn, YouTube with Peter Thiel, i.e. the same people mentioning him and me.

To compare it with some more prominent people, I repeated the same process for Hansjoerg Wyss, a prominent Swiss/American billionaire and philanthropist.

As a prominent member of the club of billionaires, Hansjoerg Wyss is much closer to his fellow billionaires Peter Thiel and Elon Musk, and also has some prominent links.

Not surprising, Forbes, which does the billionaire ranking, becomes now prominent, as well as some YouTube videos from the World Economic Forum where at times all of the people shown prominently on the map had some appearances.

This is a very short overview of a novel way of understanding somebody’s “tribe”, the context of how and where a person fits into the global social network that the Internet has become.
Ideas and feedback most welcome!