Marketing Expert's Corner

This article written in 2010
 

In 2004 an obscure book on sociology and group dynamics was published.  Instead of languishing on the shelves, this one was a hit.  The Wisdom of Crowds was not just a cult favorite -- it became a mantra repeated in countless business plans and VC cocktail parties.  Its title became the cornerstone of many -- maybe even most -- social networking sites' funding.

Surprise!  Seems like nobody actually read beyond the book's title.  My copy doesn't cover web community dynamics, and I was unable to find the word "internet" in the text at all. 

Left to their own devices, internet crowds will go stupid.  Many social networking sites are vulnerable.  Whoops. 

The Wisdom of Crowds?

You don't have to be an H.L. Mencken reader (although I do recommend Prejudices) to have heard that crowds aren't the smartest decision-makers.  It always seemed to me that the collective intelligence of committees regressed below the mean, painfully reaching the lowest common denominator.  If you believe that crowds are so smart at making choices, you'd have to think that McDonald's was America's best food.  Some smart people have said it better than me:

You can never go broke underestimating the good taste of the American public.
     -- Phineas Taylor Barnum

A person is smart.  People are dumb.
    -- Men in Black

A committee is a cul-de-sac down which ideas are lured and then quietly strangled.
     --Barnett Cocks

Nonetheless, popular culture seems to really trust crowd opinion as its guide.  American Idol makes millions of dollars a week by tapping into our popular vanity.  But the crowd here isn't too wise:  routinely, the most talented performers get booted off in favor of the cuties and hunks.  Every year, you can see Simon Cowell cringing as he has to bow before the popular votes that keep the Sanjayas in the running. 

So what is the Wisdom of Crowds thesis?  That under the right conditions, the right group of people can outperform an expert in solving problems and making predictions.  The book turns the old phrase around by explaining "why dumb people do smart things."

Boiling the book's argument down, there are three kinds of problems where crowds can be wise:  cognition (prediction, troubleshooting), coordination (sharing scarce resources, matching buyers with sellers), and cooperation (working together even when the actors are competitors -- open source is a great example).  Each of these can help businesses, and each of these can be applicable to web community dynamics.

What does it take for a crowd to be smart, rather than being a dunderheaded committee or a moronic mob?  For group judgment to really work, the members need to be:

  • independent (without authority relationships or strong influence),
  • diverse (having a range of opinion and knowledge),
  • decentralized (operating under a range of conditions and locations), and yet
  • aggregated (having a way to convert private judgments into collective decisions).  

Crowds get smart when individual errors are negatively correlated:  the dumb stuff cancels itself out, and each new crowd member is adding new information and improving decision quality.  Cool.  But how do you actually do it? 

Wise Crowds on the Internet?

The internet is a terrific way to assemble communities of interest (3 links there) -- groups that have common overall goals but diverse opinions and private agendas.  Ought to meet the Wisdom of Crowds criteria pretty well, right?

Look carefully, and you'll see that all too often it just ain't so.  eBay has done a pretty good job of aggregating and monetizing diverse, independent, and decentralized crowds.  But there are lots of examples of silly, faddish, mob-like bidding.  Like the $22,000 grilled cheese sandwich.  Exceptional, you say.  Look again:  many of the auctions settle for 90% of list price (in comparison with the best shopping sites) for an item that's clearly used.  Auctions, by their nature, drive up prices due to bad-mob effects.   Then look at the seller feedback ratings:  almost all of them are above 98% positive, even though eBay clearly has a bunch of really disreputable sellers in its community.   In my view, anyone with a rating below 99% is suspect.  But why do so few of the eBay community notice: why isn't the distribution closer to the classic 80/20 rule?  The eBay crowd is making crummy decisions.

The causes here are stem from eBay's ground rules and their  "all buyers' votes are equal" voting infrastructure.  The problem here is that the crowd hasn't been designed...merely assembled.  I won't go into the specifics of what it would take to make the crowd smarter here, but suffice it to say that the devil is in the details.

Or look at Amazon, which has spent years trying to harness smart crowds for book and product reviews.  For over a year, I've been reading the reviews extensively for products I'vebought, and guess what:  they're almost useless.  Books that are just garbage get an equal number of "it stunk!" and "best ever!" reviews.  Products with clear design and manufacturing flaws are rated five stars.  It appears that the average (arithmetic mean) Amazon score is almost useless, and the only clear signal of problems is when most votes are either one star or five stars, with few in the middle (a "bar-bell" bi-modal distribution).

The causes here are different from eBay's because votes/reviews are entirely optional and a very small percentage of buyers say anything.  The system falls into the statistical folly of self-selection.  With a high percentage of "lurkers," the results emphasize mainly extreme viewpoints.  The voices are people who are strongly impressed, or from people who know a lot or who know nearly nothing.  Again, by changing the ground rules of community interaction and adding/fixing infrastructure, the Amazon community would get a lot smarter.

Looking at the latest generation of voting communities (e.g., FaceBook, YouTube, del.icio.us, Digg), things are getting more refined.  But the inherent nature of these communities (drivenby social networks that lead to snowball effects and "mini mobs") result in storms of highly-correlated behaviors and fads.  Further, promoters and marketers are working hard to game the system and win inordinate attention.  In other words, the online mavens are trying to make the crowd dumber so they can take advantage of it.  Eventually, the crowd will tire of "the same stuff" winning over and over again, and move on to the next Hot Site.  Investors, ask your management teams what they are doing to counteract this inevitability.

Looking more broadly, the evolution of online communities seems to generally follow the law of large numbers -- the bigger the membership, the farther it goes down the bell-curve of intelligence, knowledge or taste.  For example, as the the developer community surrounding Salesforce.com has grown, it appears to have become much more basic in its competence.

How to Increase your Internet Crowd's Intelligence

Marketing is about spreading ideas.  Either you're going to tell stories that spread, or you are going to be irrelevant.
 -- Seth Godin

First, have reasonable expectations about the magic of web interactions.  Recent research at Google and MIT has shown that the wisdom of crowds and viral effects have been  exaggerated in a big way.  While it's important to find the bellwethers and influencers (à la The Tipping Point), in most circumstances, the guru is only 3-5 times more powerful than the typical community member.  Unless you're really lucky, you'll need to get the word out directly to large numbers of people before positive feedback loops develop in a meaningful way.

Second, really know who your community members are.   I recommend designing your crowd before it forms, so you can optimize the membership for what you're trying to achieve.  Before you ever start your community, develop personas, envision interactions, plan incentives, and develop a mathematical model with thresholds you can measure.  Note that conventional advertising is spectacularly ineffective in social networks -- if your business model requires ad revenues, make sure you invest the time and the energy to make the ads relevant to your community and engaging to the readership.

Third, refine the ground rules and infrastructure of community interaction.  Not all community members will be equal, and left to their own devices community debate will be dominated by first impressions, emotions, and trivialities.  Misunderstandings, misinterpretations, and ridiculous expectations based on member prejudice or ignorance tend to carry the day.  You want to make sure that your community has ways to differentiate the influence of people who are particularly knowledgeable (thought leaders) and to damp the effect of dummies (the bozos and shills).  Do not let your community fall into the Cult of the Amateur effect -- where worthless opinions and ego blasts shout out individuals with better judgment, wiping out the wisdom effects of a healthy community.

Further, design community voting systems to avoid easy manipulation.  From flame wars to Google-bombing to Digg-rigging, individuals and companies know how to override bona fide community sentiment.  Maintaining an automatic credibility index for community members is an important advance -- bringing the Delphi method to real time interactions.

When you build out your community site, either create your own infrastructure or use tools from companies like BuzzMetrics, VML, BazaarVoice and others to monitor and optimize community interaction. 

Finally, the healthiest communities seem to involve some human intervention and moderation.  While the need for this is ironic in the online world, communities seem to behave better and smarter when members know they aren't really anonymous, that someone is looking, and that someone cares about the quality of the outcome.

Phone Us +1 650 326 2626