Network Effects Are Overrated

Netflix provides a stark example of the tendency to exaggerate the role of network effects and AI.


Continue reading the main story

Supported by

Continue reading the main story

The DealBook newsletter delves into a single topic or theme every weekend, providing reporting and analysis that offer a better understanding of an important issue in the news. If you don’t already receive the daily newsletter, sign up here.


The recent explosion in value of technology stocks has entrenched dangerous beliefs about what qualities characterize the best businesses. These had been gaining currency over the last decade as a handful of so-called platform companies came to represent a large share of the overall market’s collective value. In place of old-fashioned forms of competitive advantage, new superpowers are attributed to these platform businesses, often to justify otherwise inexplicable valuation metrics.

What platforms have in common is that their basic value proposition lies in the connections they enable, whether by encouraging innovation, facilitating transactions, or deepening relationships. Their imagined invincibility derives from a combination of unstoppable network effects and futuristic artificial intelligence. The idea is that the internet has expanded the scope of these networks and the enormous quantities of data they generate feed new life into the old A.I. algorithms.

Sometimes also referred to as the “flywheel effect,” network effects occur when every new user increases the value of the network to existing users. In a digital environment, it is argued, not only do new users attract still more new users ad infinitum, but the continuous improvements facilitated by A.I. make the prospect of successful competitive attack ever more remote, leading inexorably to a world dominated by impregnable winner-take-all markets.

The problem with this narrative is that it ignores the numerous ways in which the new digital platforms actually make businesses more vulnerable to competitive attack compared with the analog models that they have disrupted. The ease with which customers can switch undermines captivity and the asset-light nature of these businesses both lowers entry barriers and the level of activity required to break even.

Compare online with offline retail malls for example. Traditional malls had two major business advantages: Their vendors were committed to long-term leases and their shoppers’ next best option was many miles away. By contrast, on the internet, alternatives for buyers are only a click away and sellers offer their wares across competing platforms. The point isn’t that you would rather invest in a mall operator than Amazon during a pandemic but simply that offline business models have surprising relative resilience. Despite the secular trends, right up until the Covid-19 crisis hit, struggling online retailers were increasingly looking to solve their structural woes by opening up mall outlets.

Even among the companies that have come to define the sector — Facebook, Amazon, Apple, Netflix and Google — only Facebook’s franchise was primarily built on network effects. Yet all these companies seem to appreciate the public relations and investor relations value of pushing the platform folklore. Internal email correspondence unearthed by congressional investigators revealed that their C.E.O.s understand both the extent of their vulnerability and the value of the mythology. Confronted with the challenge of explaining away various competitive threats, Facebook’s vice president for corporate finance and business planning proposed a solution: “We need a simpler ‘platform’ story.”

The case of Netflix provides a stark example of the tendency to exaggerate the role of network effects and A.I. to justify outsize valuations. When it sold online subscriptions for DVD rentals, its competitive advantages came from the same two primary sources that most strong franchises have historically relied on — scale and customer captivity. The scale benefits came from spreading its fixed costs of marketing and maintaining its national distribution infrastructure across its larger user base. The captivity came from the addictive service and powerful recommendation engine.

When the model shifted toward streaming starting in 2008, the sources of advantage remained the same although the fixed costs of the physical distribution were now replaced with the fixed costs of digital storage and distribution. Conventional wisdom, however, is that the Netflix streaming model has unleashed a variety of supernatural powers flowing from A.I. and network effects. The result, the Deutsche Bank analyst Bryan Kraft wrote in a 2019 research report, has been to bestow upon Netflix a mystical “platform status,” implying stronger entry barriers and supporting higher valuations. Such true believers think for instance that the streaming model not only allows Netflix to refine its already excellent recommendation engine, which it does, but actually bestows on Netflix a magical ability to algorithmically pick hits, which it does not.

This particular canard begins with the origin story behind Netflix’s first big hit, “House of Cards.” As The Times columnist David Carr described it, Netflix was able to prudently outbid all other comers for two seasons of the series — 26 episodes in total for a reported $100 million — without so much as a pilot because of structural advantages bestowed by big data and artificial intelligence. In this telling, competitors were not privy to three key bits of data that together made “House of Cards” a surefire hit: the popularity of films directed by David Fincher, films starring Kevin Spacey, and the original BBC “House of Cards” series with Netflix viewers. “With those three circles of interest,” Mr. Carr wrote, “Netflix was able to find a Venn diagram intersection that suggested buying the series would be a very good bet.”

Such ex-post explanations for the selection of successful creative projects suggest a false level of predictability. They inevitably follow hits just as deafening silence follows flops. Soon after the triumph of “House of Cards,” Netflix committed to an even more expensive series — “Marco Polo.” Dropped by the original buyer, Starz, because of the prohibitive expense and complications of filming in China, the first two 10-episode seasons had an estimated budget of $180 million. When the show was canceled, no suggestion of an algorithmic glitch was provided.

Increased original content spending has been the most significant change since then in the Netflix business model, and this reflects not a better business but increased competition from companies like Disney, WarnerMedia, ViacomCBS and NBCUniversal upon whom Netflix relied for its licensed content. “Reading a script and guessing who might be good to cast in it — it’s not something that fundamentally as a tech company … we’re likely to build a distinctive organizational competence in,” the Netflix C.E.O., Reed Hastings, told Fast Company around the time of the original “House of Cards” investment. His conclusion could not have been clearer: “We think that we’re better off letting other people take creative risks.”

Netflix’s unleashing the Kraken of content investment reflects competitive necessity not newly discovered competitive advantages. Of course, the easiest way to check whether barriers to entry have gone up or down is to see how much entry has occurred since. Just between the beginning of 2019 and the end of 2020, Netflix went from representing almost half of U.S. subscriptions for on-demand video services to around a quarter, according to data compiled by The Wall Street Journal.

For almost 20 years, Netflix has repeatedly tried to incorporate network effects into its core business model. But it has failed consistently, and ultimately given up.

Even going back to the days of DVD by mail, Netflix tried to create its own form of social networking by establishing Netflix Friends in 2004. Despite never gaining traction, the company held on to the service until 2010 before shutting it down. A number of subsequent programs with Facebook, one of which Mark Zuckerberg personally had a hand in designing, were also discontinued for lack of user interest. Netflix even eliminated user reviews altogether in 2018. Hastings himself ultimately described his futile quest for network effects to me as a “competitive fantasy.”

Netflix is only expected to break even on a cash flow basis this year after having collectively accrued around $10 billion in negative cash flow since it began investing in original content in 2012. Many analysts justify their valuations of streaming services based on a belief that at scale these services will achieve the same profit margins as the cable channels they are increasingly replacing.

Don’t bet on it. Cable channel customers are predominantly distribution businesses offering long-term contracts in an environment with capacity constraints that limited competition. Streaming service customers churn relentlessly and have few barriers to entry beyond a big enough checkbook. And with at least half a dozen deep-pocketed competitors bidding against one another for content and talent that might drive subscribers, I would rather be Shonda Rhimes or the owner of some proven intellectual property.

The true danger of believing the various myths of the platform era lies in what happens when market euphoria subsides, as it inevitably does, and making informed distinctions among technology investments becomes essential for financial survival. Whether investing, operating or regulating, it is critical to pinpoint the true source and extent of individual advantage rather than relying on simplistic delusions to guide decision-making.

Mr. Knee is the author of “The Platform Delusion: Who Wins and Who Loses in the Age of Tech Titans.” He is a professor of professional practice at Columbia Business School and a senior adviser at Evercore.


What do you think? Let us know:

Leave a Reply