Is Google's de-emphasis of its social network proof of Facebook's competitive advantage?
While Facebook (NASDAQ:FB) is worth more than $270 billion today, its business is still a fairly new breed. Founded in 2004, the company quickly grew to become one of the world's first major online social networks. But many questioned whether the company's business was sustainable. After all, online social networks hadn't yet proved to be enduring business models. And even after the company's high-profile initial public offering in 2012, there were still a handful of critics who believed Facebook was lacking a sustainable competitive advantage. Today, however, it's clearer than ever that Facebook's competitive advantage is both powerful and sustainable.
What Is Facebook's Competitive Advantage?
Long-term investors are often searching for potential investments with durable competitive advantages. As famed investor Warren Buffett says, an investment should have a "moat."
Buffett has said.
"In business, I look for economic castles protected by unbreachable moats."
After all, without a competitive advantage, how can investors expect their current investment to be around years into the future, let alone appreciate in value?
But what is Facebook's moat? Investors have argued that its competitive advantage is a network effect. A network effect becomes a competitive advantage when each additional member in the market makes a network stronger for all parties involved, and when, on the flipside, users are in a potentially disadvantageous situation when they aren't a part of the network.
Consider Facebook's network effect. With about one out of seven people in the world using the service on a monthly basis, there is a high probability that users who join the service will have friends and family already actively using the service. So for every additional user, the network becomes increasingly useful, making it easier for existing members to connect. On the other hand, if someone chooses to leave Facebook's network and use a similar yet less established social network, that person will have to leave behind many friends and family who use only Facebook. A network effect, therefore, makes it very difficult for new networks to woo users away from leading networks.
Perhaps the best illustration of the strength of Facebook's social network has been Google's failed attempt to compete with Zuckerberg & Co.
Google+: Far From A Threat
Hoping its sheer size and reach through Google search and Google products such as Gmail and YouTube would launch Google Plus into a position to eventually compete head-to-head with Facebook, Google initially required a Google+ account to engage with others on its products.
The idea was that Google+ would be a "platform layer that unified Google's sharing models." Even more, Google+ was aimed at becoming its own "product/stream/app in its own right," said Google executive Bradley Horowitz in a Monday post on Google+ detailing Google's decision to phase out the importance of Google+ as a platform across its products. Horowitz took over Google+ earlier this year.
"This was a well-intentioned goal, but as realized it led to some product experiences that users sometimes found confusing. For instance, and perhaps most controversially, integration with YouTube implied that leaving a comment on YouTube (something users had obviously been doing successfully for years) suddenly and unexpectedly required "joining Google+."
Horowitz said Monday in a separate post on Google's official blog.
"So in the coming months, a Google Account will be all you'll need to share content, communicate with contacts, create a YouTube channel and more, all across Google."
The de-emphasis of Google+ as a platform across Google's products, as well as the clear lack of traction the platform has had in the past few years, highlights just how difficult it is to compete with Facebook's network effect. And if Google can't compete effectively with Facebook, who can?
Here’s an infographic that nicely illustrates the value of users and their impact on the network effect.
Click Image To Enlarge
Metcalfe's Law of Network Effects vs Zipf's Law
One of the surprising things about the valuation of Facebook in its IPO back on May 2012 is how the rationalization for the $104 billion valuation seems to have been based on some varient of Metcalfe’s law of network effects. Anthony Wong Kosner, a contributing editor for Forbes said in an article dated May 31, 2012 said,
"Surprising, I say, because this law (aka Metcalfe's Law), that states that the value of a network (network effect) is proportional to the square of the number of the nodes on the network, has been refuted quite convincingly as a method of evaluating social networks at least since Andrew Odlyzko and Benjamin Tilly’s groundbreaking paper in 2005 and the more widely cited article with Bob Briscoe in IEEE Spectrum in 2006. Their candidate for modeling value? The less sensationalist Zipf’s law."
Kosner goes on to say in the above referenced article:
And yet, as recently as the day before the IPO, Tim Beyers of the Motley Fool defended the impending valuation on AOL’s Daily Finance channel on the basis of Metcalfe’s law. Consider that five years ago, Forbes Editor William Baldwin picked up the baton from Odlyzko and wrote in his column,
“Metcalfe’s Law is a very optimistic statement about connectedness. Double the number of nodes in a network—of computers, phones, rail stations or, perhaps, friends—and you quadruple its value. Too optimistic, says a revisionist mathematician, Andrew Odlyzko. His measure of networking value comes from a different law, described 75 years ago by the linguist George Zipf.”
What is Zipf’s law? It is one of a family of related discrete power law probability distributions that describe how in any system of resources there are a small number of items of very high frequency and value and a “long tail” of many more of decreasing frequency and value. It is, in essence, an empirical description of hierarchical distribution of resources (i.e., rich get richer…). Zipf discovered the law through the observation of the frequency of words in the English language, but the law holds even for randomly generated “languages,” as well. The law states that,
“Given some corpus of natural language utterances, the frequency of any word is inversely proportional to its rank in the frequency table. Thus the most frequent word will occur approximately twice as often as the second most frequent word, three times as often as the third most frequent word, etc.”
Social networks are, in fact, more like languages than they are like the ethernet networks that R0bert Metcalfe based his law on. The difference is that the nodes of hardware networks are fairly homogenous. Each one has an equal relationship to every other one. But social networks are anything but homogenous. Dunbar’s number sets the size of a meaningful social group for humans at about 150. Beyond that size, additional members add diminishing value. But even within that 150 there is considerable range in affinity and sentiment.
Ben Kunz wrote about this recently in Businessweek in terms of the threat to Facebook of the rise of what he calls “unsocial” networks. He points to Path, which limits your network to Dunbar’s 150, and a new app for networks of only two, called Pair. These new micro social networks are bound to get closer to what their users really care about than sprawling sites like Facebook, but will probably be even more impenetrable to advertisers, which will make monetization difficult.
An example of the difference in valuation between Metcalfe’s law (n2) and Zipf’s law (n log (n)) is illustrated in that 2006 IEEE article:
"Imagine a network of 100,000 members that we know brings in $1 million. We have to know this starting point in advance—none of the laws can help here, as they tell us only about growth. So if the network doubles its membership to 200,000, Metcalfe’s Law says its value grows by (200,0002/100,0002) times $1 milion, quadrupling to $4 million, whereas the n log( n ) law says its value grows by 200,000 log(200,000)/100,000 log(100 ,000) times to only $2.1 million.”
To figure out what Facebook is worth based on this rationale would involve having a starting point valuation, and the first public milestone was Microsoft’s October, 2007 purchase of a 1.6% share of the company for $240 million, giving Facebook a total implied value of around $15 billion. This was clearly a wild overvaluation, considering Yahoo and others were offering $1-2 billion to buy Facebook outright at the same time. But just as an illustration of relative value, let’s call it $1 billion in 2007, and take Facebook’s 50 million monthly active users in October 2007 and its 900 million now. Based on the $1 billion valuation in 2007, the Metcalfe method would yield a valuation of a whopping $325 billion. The Zipf method would yield a much more restrained valuation of $21 billion. Further research has suggested that the “actual value” of a social network is somewhere between the Metcalfe and the Zipf numbers, with the growth line tending more linear than exponential.
But the difference in value between these two approaches is not just about market valuation, but, more importantly, about value of the networks to the users. Zipf’s law would suggest that most of what is coming through the Facebook pipe to users is of little value to them. But by another important metric, Facebook is clearly very valuable to its users. In an under-read but extremely valuable story this weekend on Forbes.com, contributor Nir Eyal talked about what he referred to as the most important “hacker metric.” Referring to a post by his friend Andrew Chen, Eyal says that the fact that “Facebook’s historical ratio of daily active users (DAU) to monthly active users (MAU)” has consistently been about 50% is highly significant.
Daily active users are, in Eyal’s terms, the agents of a successful startup. The timeframe is important because of what is called “Viral Cycle time,” or the amount of time it takes to complete a viral loop.
“Having a greater proportion of DAUs dramatically increases Viral Cycle Time for two reasons. First, daily users initiate loops more often—think tagging a photo on Facebook. Second, more daily active users means more people to respond and react to each invitation. The cycle not only perpetuates; with high DAU to MAU, it accelerates.”
Eyal contrasts the micro-engagements of users with the performance of the overall network,
“As opposed to distribution channels, the mechanics driving user engagement do not follow a power law. In fact, it is the nuances of user behavior that make the competition irrelevant, just as it did in the case of Facebook’s early rivals.”
What Chen and Eyal are talking about are, in effect, quantum effects. But these viral effects do not scale up exponentially into the value of the entire network. The network encounters material resistance that act as a curb on exponential growth.
Perhaps Facebook and its evaluators have been so enthralled by the levels of user engagement that they have overlooked the way the value to individual users diminishes beyond their core network. This may explain why Facebook’s vast numbers do not translate into call-to-action advertising success. On Facebook we are actively engaged with what we most care about and we tune out the rest. The rest may be more than half of the content, and the vast majority of the marketing messages, but we are so in need of the wheat that it takes a lot of chaff to diminish our satisfaction. Beyond a certain point, though, the growth of the size of ones personal or Facebook’s global network does not increase a user’s happiness. It’s like the studies that have determined that over a certain threshold more money does not make people substantively happier.
Once you accept the fact of how differentiated the different members of a user’s network are in terms of their value and relationship to the user, the economies of scale of Facebook kind of break down. It is not a matter of general social clout (or Klout as the case may be) but of the specific user’s value overlay on their own network. If an advertiser really wants to target a specific type of user, given this complexity, wouldn’t it be easier to take out a quill pen and inscribe a missive on a piece of letterpress stationary and hand deliver it by horse-drawn carriage? At least that would get their attention!
Reverse Network Effects
NETWORK EFFECTS ARE the holy grail for Internet startups looking for venture-scale returns. On a platform with network effects, the value to a user increases as more users use it. Facebook, Twitter, LinkedIn, YouTube, Skype, WhatsApp and many others benefit from this dynamic.
In an age when more than a billion people connect over a network and new networks reach multi-billion dollar valuations with a handful of employees, one is tempted to believe that online networks are almost fail-proof. But as online networks grow to a size never seen before, many question their sustainability and believe that they are becoming too large to be useful.
To explore the future of online networks, it’s important to note how network effects correlate with value and the factors that make these network effects work in reverse.
Network Effects and Value
There is a strong correlation between scale and value in businesses with network effects.Greater scale leads to greater value for users, which in turn attracts other users and further increases scale. This rich-becomes-richer dynamic allows networks to scale rapidly once network effects set in.
There are three sources of value created on networks: Connection, Content and Clout.
Connection: Networks allow users to discover and/or connect with other users. As more users join the network, there is greater value for every individual user. Skype and WhatsApp become more useful as a user’s connections increase. Match.com and LinkedIn become more useful as more users come on board.
Content: Users discover and consume content created by other users on the network. As more users come on board, the corpus of content scales, leading to greater value for the user base. Content platforms like YouTube, Flickr and Quora, as well as marketplaces like AirBnB and Etsy becomes more useful as the number of creators and the volume of content increase.
Clout: Some networks have power users, who enjoy influence and clout on the network. Follower counts (Twitter), leaderboards (Foursquare) and reputation platforms (Yahoo Answers) are used to separate power users from the rest. On networks like Twitter, the larger the network, the larger is the following that a power user can develop.
Across these three drivers, a network with greater scale provides greater value in the form of:
- More prospective connections for the user
- A larger corpus of potentially relevant content
- Access to a larger base of potential followers (greater clout), for power users
On most networks, value for users is created through more than one of these three sources. Facebook, for example, started with a value proposition centered around connection, but the introduction of the news feed has made content a central driver of value. In recent times, the addition of the subscribers feature has added clout for some Facebook users as well.
Why Network Effects Work in Reverse
One would expect that the bigger the network, the more value users derive from it.
However, as networks scale, the value for users may drop for several reasons:
- Connection: New users joining the online community may lower the quality of interactions and increase noise/spam through unsolicited connection requests.
- Content: The network may fail to manage the abundance of content created on it and may fail to scale the curation of content created and the personalization of the content served to users.
- Clout: The network may get inadvertently biased towards early users and promote them over users who join later.
Just as network effects create a rich-becomes-richer cycle leading to rapid growth of the network, reverse network effects can work in the opposite direction, leading to users quitting the network in droves. Friendster, MySpace and Orkut bear testimony to the destructive power that reverse network effects wield.
Reverse Network Effects: Connection
Connection-first networks (dating websites like Match.com and networking communities like LinkedIn) build value by connecting people.
These networks may suffer from reverse network effects as they scale if new users joining the network lower the value for existing users. To prevent this, an appropriate level of friction needs to be created, either at the point of access or when users try to connect with other users.
On dating sites, women often complain of online stalking, as the community grows, and abandon the site. Sites like CupidCurated have tried to solve this problem by curating the men that enter the system, in a manner similar to restricted access at a singles bar.
LinkedIn creates friction by preventing users from communicating with distant connections. This ensures that users do not receive unsolicited messages. This also allows LinkedIn to offer frictionless access (OpenMail) as a premium value proposition.
ChatRoulette, in contrast, anonymously connects users over a video chat without needing to login. This lack of friction led to ChatRoulette’s stellar growth but also led to reverse network effects as anonymous naked hairy men took to the network, thus increasing noise and driving genuine users away from it.
Dating sites, as well as social networks like Orkut, have imploded in a similar manner after reaching scale, owing to noise created by fake profiles.
In general, networks of connection scale well when they create appropriate barriers to access on the network.
Reverse Network Effects: Content
On content networks like YouTube or Flickr, a larger network is likely to have more content creators, leading to more content for the user to consume. Networks like Facebook and Twitter, in addition to being networks of connections, are also networks of content.
Most networks of content have low friction in content creation to encourage activity from users and reach critical mass faster. To ensure that the content is relevant and valuable, the network needs strong content curation and personalization of the user experience.
Reverse network effects set in if the content curation systems don’t scale well. As more producers create more content, the relevance of the content served to consumers on the network shouldn’t decrease.
Content networks create a curation mechanism through a combination of moderation, algorithms and community-driven tools (voting, rating, reporting etc. ). Voting on YouTube, flagging a post on Facebook and rating on Yelp are examples of curation tools.
Curation mechanisms often break down as the volume of content increases. When curation algorithms and moderation processes do not scale, noise on the system increases. This leads to reverse network effects and users abandoning the system.
Quora has a very strong curation mechanism in place and benefits from a tech-savvy early user base. As Quora scales, many worry that less sophisticated users, entering the system, may increase noise leading to a rapid depletion of value for existing users. It remains to be seen whether its curation can scale as the network opens up to a broader user base.
Content networks need a personalized consumption experience for users, that serves them relevant content.
An example is the news feed on Facebook or Quora or the recommendation system on YouTube.
Inability to maintain relevance of the consumption experience, with scale, may create reverse network effects.
The user experience on Facebook is centered around the News Feed. However, Facebook’s frictionless sharing and cluttered news feed may lead to lower relevance for users as the network scales. Several factors contribute to this:
- When a user adds friends indiscriminately, her news feed becomes cluttered with irrelevant posts.
- Noise is further increased when marketers and app developers get access to the news feed.
- When networks like Facebook and Twitter implement monetization models like Promoted Posts/Tweets, the signal to noise ratio suffers further as promoted content is less relevant than organic content.
Networks of content are constantly faced with the risk of reverse network effects as they scale. The poor signal-to-noise ratio in the news feed, not the size of the overall network, is Facebook’s weakest link as the network scales.
Reverse Network Effects: Clout
Networks of clout have a system of differentiating power users from the rest. Twitter, Quora and Quibb have baked in clout through the one-sided follower model. Active users vie for greater glory while using the network.
Networks of clout tend to be biased against users joining in late. Clout is a consequence of content that the user creates and early users get more time to create content and develop a following.
This is, ironically, aggravated by focusing on a high signal-to-noise ratio. Twitter recommends super users to prospective followers as these users are likely to create better content. Hence, the platform itself helps separate the power users from the rest.
Users who join later find it more difficult to develop a following and may stop using the network. These networks need a mechanism to ensure new users have equal access and exposure to the community to develop network clout. 500px, for example, differentiates Top creations from Upcoming creations to expose recent activity (often from undiscovered users) to the community.
Reverse network effects often cause a large and thriving network to implode. As a network scales, it’s ability to maintain a high signal-to-noise ratio is the leading indicator of its usefulness. Networks can, in fact, scale very well and prevent reverse network effects from setting in if they have
- Appropriate level of friction in network access and usage, that prevents abuse
- A strong curation system that scales well with the size of the network
- A highly relevant and personalized user experience
- A democratic model for users to build influence
Networks that have excelled in the above have scaled well. In a world where networks are reaching unprecedented scale, a keen focus on maintaining a high signal-to-noise ratio will enable them to remain valuable and effective as they grow.
Courtesy of an article dated July 28, 2015 appearing in MotleyFool.com, an article dated June 8, 2013 appearing in Pamorama.net, an article appearing in the March 2014 issue of WIRED, and an article dated May 31, 2012 appearing in Forbes Tech