2014-10-July-Train-Signal-Noise.jpgSocial information is very different from traditional documentation. We've grown accustomed over the past few years to referring to documents as “unstructured data.” But even PowerPoint is neat and orderly compared with the wild frontier of enterprise social.

Hashtags, "likes" and abbreviations are all part of users conveying information. Almost all organizations with enterprise social deployments turn to it after traditional documentation, but in many cases, social knowledge management is an afterthought -- if it's performed at all.  Yet information theory shows us that enterprise social is much more likely to contain meaningful enterprise knowledge than traditional documentation. 

How should we approach knowledge governance in social streams?

The World Wonders

James Gleick, in “The Information” revisits the classic Marshall McLuhan medium/message conundrum for the internet era.

Information Theory teaches us that almost all information carries extraneous clutter, background noise, padding, carrier waves, etc., that are unrelated to the core “message.” The earliest days of analog signaling rapidly uncovered the signal to noise problem -- if a message is hard to hear because of background static, turning up the volume amplifies the noise along with the signal.

Digital signals are not immune from signal/noise issues. During the Battle of Leyte Gulf in October 1944, US Navy Pacific Command lost track of the fleet group Task 34 commanded by Admiral Halsey when he chose to pursue a decoy Japanese fleet instead of following an order to protect an Allied amphibious landing force. In those days, coded messages also included extra words that had nothing to do with the core message. As a result, the encrypted message listed below included extra padding at the end. (The actual message is in bold.)


Unfortunately, when the message was decrypted, the corpsman handed the Admiral a message that probably generated even more noise.



Task Force 34, Located? © 2014 Christopher F. McNulty

The actual information in the message -- where is the fleet -- was contained in a lot of extraneous data. (For more, see The World Wonders). The point? It can be hard to distinguish signals from the noise.

And we don’t need as much detail as we think to convey a message. Consider the two following “signals”:

  • At the June meeting, the committee established November 10, 2015 as the end date for the Montana project. (92 characters)
  • @ Jun mtg - cmte set 11/10/15 4 Montana prjct end (49 characters)

Both convey the same information. The second message is more succinct, and probably has a higher signal to noise ratio. However, a verbose message may be more easily read -- and a terse message is more susceptible to misunderstanding e.g.,  is 11/10/15 November 10, 2015 -- or October 11, 2015?

During World War II, Claude Shannon focused on the likelihood, or unexpectedness, of a message as an expression of its information carrying capacity:


A simple example? In English, the letter q is almost always followed by u. So the u conveys almost no actual information and could be considered part of the noise -- not the true information. Even if I leave out the second letter of the airline:


You can probably still understand the name.

Social Signals, Social Noise

Like the Pacific Fleet, we have well practiced habits of “padding” our documentations with logos, headers, explanations and properly written sentences. However, the “nut” of a given document might be a simple fact -- e.g., the committee approved the project because the chairperson believed it would reduce operating expenses. This may however be buried in an executive summary or at the end of meeting minutes.

Managing enterprise social information was viewed to be a simple undertaking. Earlier in my career, we debated the pros and cons of social migration tools. It had been thought that enterprise social messages were transient parts of a collaboration process that might be purged in a few months -- all you ever needed to retain were the end documents.

This approach proved short sighted. First, as many of the first wave of collegiate Facebook users discovered after 2004, the Internet has a long memory. Things you post tend to endure longer than anyone expected. Enterprise social is no different.

More importantly, the actual information needed to answer a future question -- Why did we do this project anyway? -- may no longer be found in a document. It may live solely in a Yammer post, for example. If unpredictability increases the potential information level, watch out. Social posts are highly unpredictable.

So how do you control social information? Just remember the three C’s:

  • Curate -- If your enterprise social streams contain intellectual property and unique elements of knowledge management, some level or review and retention helps remove extraneous or erroneous posts.
  • Classify -- Documents were already “unstructured” -- but they are often based on repetitive templates and use full text grammar. Social posts can be really hard to find if user spellings and syntax adopt Twitter conventions. Metadata tags and classification bridge the usability gap if users are going to sustain the information over the long haul.
  • Context -- Keeping documents close to the conversations -- and keeping the conversation close to the documents -- helps assure a 360 degree view of the necessary information. If embedded links point to obsolete document libraries, or posts are purged indiscriminately after a year, you’re reducing the signal and just leaving behind the noise.

Social communications are going to contains a lot more mission critical enterprise information than you may have anticipated. Information without meaning or purpose is just data. Be ready! The world wonders.

Title image by Paolo Bona / Shutterstock.com