Network effect in reverse?

"When the Network Effect Goes Into Reverse" is the title of a James Stewart article from last Friday in the NYTimes. It caught my attention because I'd never heard of a network effect reversing. At first blush, I wasn't even sure what that meant.

I'm not certain if there's a canonical definition of network effects, but the one I've always worked from is from Hal Varian's Information Rules (if you work in the internet space and haven't read this classic, pick it up posthaste) which seemed like a generalization of the law named after Bob Metcalfe, which was specific to telecommunications networks.

Metcalfe's Law, per Wikipedia (which gives discovery credit to George Gilder):

The value of a telecommunications network is proportional to the square of the number of connected users of the system.

Hal Varian on network effects, from page 13 of Information Rules:

When the value of a product to one user depends on how many other users there are, economists say that this product exhibits network externalities, or network effects.

If you read Stewart's article, he seems to reach for a literal reading. That is, if you start losing users, the value of the product or service also declines. Stewart quotes internet analyst Ken Sena:

“The network effect allowed these companies to grow so fast, but the decline can be just as ferocious,” Mr. Sena said. “If any of them misstep with users, they can leave, and the network effect goes into reverse.”

There's the germ of an interesting idea in Sena's quote, but the rest of Stewart's article doesn't tease it out. Instead, the examples Stewart goes on to cite fail to support his headline, and many are just plain false. He writes::

Facebook has been a classic example. If your friends, colleagues or classmates are all on it, you’re all but compelled to join. But evidence that the network effect is working requires rapid growth in users and revenue, especially during the early stages of a company’s public life. So far, social media has failed to deliver the kind of growth that would bolster investor optimism, let alone euphoria.

How is Facebook not one of the most canonical examples of the power of the network effect? Stewart's point seems to be that since Facebook's user growth has slowed since they went public, the network effect is failing? Facebook has over 900 million users! Despite the law of large numbers, it's still growing.

To me, the only growth problem that Stewart says social media needs to deliver to "bolster investor optimism" is monetization. In terms of user growth, I'd be hard pressed to think of many faster growing companies in history than recent social media examples like Facebook, Twitter, Instagram.

Then Stewart shifts to Groupon:

This week’s Groupon earnings illustrated the problem for social media companies. In theory, Groupon should benefit from the network effect. The more users it attracts, the more merchants will want to offer coupons through Groupon, and vice versa. And on the face of it, the earnings report looked good. Groupon earned a profit of $28.4 million for the quarter, above analysts’ expectations, reversing a loss a year ago. Groupon’s boyish-looking chief executive, Andrew Mason, called it a “solid quarter.”

But growth, not profit, is what matters at the early stage in the life of a networked Internet company. Groupon said revenue grew 45 percent over the same quarter last year. But it counts payments that it passes on to merchants as part of its revenue. When that part of revenue was excluded, revenue grew just 30 percent. And compared with the previous quarter, revenue grew just 1.6 percent.

First of all, Groupon is not really a network effect company. Remember Varian's or Metcalfe's definitions above. The direct value of Groupon to me hasn't changed appreciably even though its user base has grown. Any company with a lot of users will be attractive to advertisers or merchants, not just Groupon. That is unrelated to network effects. I haven't noticed any appreciable difference between Groupon when it had many fewer users and today, when it has many more users. [1] But from the early days of Facebook and Twitter to where they are today, the value of the service to me has shifted dramatically from its now massive user bases. Those are network effect companies.

Later Stewart writes that "a positive network effect is supposed to exclude competitors, but Groupon has long suffered from the perception that it's vulnerable to competition." He actually answers his own question, because Groupon is not a network effects company, and thus it doesn't benefit from the competitive protection such a model would offer.

Secondly, Stewart conveniently looks at only the last two quarters of Groupon's results and refers to those as "early stage". In its actual "early stages", Groupon was, by many accounts, the fastest growing company in history, reaching massive revenue milestones in times that I don't think would have been possible prior to the invention of the internet. Stewart's implication that Groupon didn't have enough growth at its early stages is just wrong. Groupon does have issues,  but a reversal of network effects is not it. 

So if this article doesn't support the concept of reversed network effects, then what does? If we go back to Varian's definition, there is another way to interpret "network effects in reverse," and that is by reversing the value of every additional user. That is, "network effects in reverse" could mean that a the value of a product or service decreases with each successive user. In fact, whether we call it "network effects in reverse" or something else, perhaps "network defects," there are many examples of companies that reach a state where such a phenomenon occurs, and I believe it can explain why some internet companies can spiral into rapid decline in such a sudden fashion.

It's a topic that interests me quite a bit, so I'll cover in in another post.

1 Groupon had one very light network effect built into its product early on in that certain deals wouldn't convert unless enough people purchased it. So having more users on the network could affect the rate at which deals I purchased became valid. But for a long time now, it seems as if essentially every GroupOn deal converts. I'm not a heavy user so my recollection of this is fuzzy.