The Platform Delusion: Who Wins and Who Loses in the Age of Tech Titans

By Jonathan Knee, 2021 (384p.)

This book derives from a class called Digital Investing that professor Knee teaches at Columbia Business School since 2015. It is the most insightful book I have ever read on technology investing. It is even better and more relevant than Platform Revolution (2016). Knee is a highly respected authority on the topic and his knowledge draws from decades of experience in investment banking and consulting in the technology and media spaces.

What makes this book so different is that Knee has a strong and bold opinion on most of the key players in the Big Tech space.  I forgive him for not mentioning Upstart at all, and for mentioning Accenture only once, but he’s so generous otherwise, my alma mater is lucky to have him. Chapter 8, where he gives the reasons for his confidence in Google’s dominance whatever happens, was my favorite. Instead of commenting here on his many theories and conclusions about winners and losers in the technology space (from internet aggregators, to content creators, Software as a Service (SaaS) providers, cable owners, and trading platforms), I comment selectively in the highlighted passages below.  At the very end of this review I also include a link to the only recent YouTube video I found with Professor Knee, as well as the Amazon description of each of the last three books he published before Platform Delusion.

If I could summarize Jonathan Knee’s view in one line, here is what it would be:  Buy Google, Visa,, and Cable companies – sell Facebook, Netflix, and Apple – short ad agencies and any ad-tech company that competes with Google.


PS:  There is another outstanding book I recently read that comes close to being just as good as Platform Delusion.  It’s called The Future of Money: How the Digital Revolution Is Transforming Currencies and Finance, and it also came out last month. The author is a professor at Cornell and the book is unlike any other on the topic.  It is much better and more relevant than Ken Rogoff’s The Curse of Cash (2016).  …when it rains [good books], it pours.

HIGHLIGHTED EXCERPTS (AA Comments in italics)

Part I: Digital Advantage and \disadvantage

Chapter 1: The Four Pillars of the Platform Delusion

Even Zoom, the video communications platform that is perhaps the most iconic success of the pandemic era, is not really a strong network effects business. By the end of 2020, the company was worth over $100 billion and its stock traded at around ten times its 2019 offering price. Zoom is a fabulous product, but its very success in eliminating any friction or complexity in adoption has severely limited how powerful its network effects can be. [AA Comment: I still use Zoom. But I’m a dinosaur. Not important who delivers it as long as it works.]

More broadly, the authors of The Business of Platforms [AA Comment: published May 7, 2019], looked at two decades of performance, from 1995 to 2015, of platform businesses and noted that relatively few had survived. Specifically, “only 17 percent (43 out of 252) remained in 2015 as independent public companies.” [AA Comment: 17 is a special number.]

“In a relatively small number of platform spaces, failure was a function of a genuine winner-take-all or winner-take-most outcome for a competitor.” [AA Comment: That’s nature. It’s like saying: Only in a relatively small number of cases does a species survive evolution. The Pareto principle (a.k.a. 80/20 rule) has been out since Vilfredo Pareto described it in 1906.  A relatively small number of companies (i.e. 17%), account for most (i.e. 83%) of the consequence (i.e. value) in nature.]

What is alarming, however, is the extent to which the euphoria triggered by the Platform Delusion has led investors to forget that, ultimately, the existence of competitive advantage is what drives the ability of any business, digital or analog, to produce consistently superior returns. [AA Comment: easy to say in retrospect and could have been said in 2015 when this course material was being assembled. I wonder for how long the professor has been bearish. In my opinion it is better to stay humble here.  That said, I like his style. I also like how well the book was referenced and researched.]

Although frequently presented as peculiar to the internet, platform business models have long been ubiquitous. [AA Comment: This is so true.  Think Sherwin Williams, O’Reilly, Old Dominion, Goldman Sachs, American Express, the Medici, Fugger, Guggenheim, Harvard, the Vatican, the Evil Axis, and Marcel Duchamp.]

Chapter 2: Network Defects: Scale in the Digital Era

…it is possible to identify the maximum number of profitable competitors. If a particular sector opportunity can sustain a competitor at a 5 percent market share, twenty market participants can thrive indefinitely. If high fixed costs dictate a 35 percent break-even market share by contrast, monopoly or duopoly are the only sustainable market structures. [AA Comment: Cool that he puts numbers on it.]

So, in ride sharing, where the ability to deliver a car within three to five minutes dominates all other customer considerations, adding drivers to the network beyond this point is of little value.

The insurance industry created a nonprofit called Insurance Services Office (ISO) fifty years ago to pool data from their property/casualty insurance members to improve their collective risk assessments. The business became a for-profit in 1996 and today represents the core asset of Verisk Analytics, a $30 billion public company. [AA Comment: Verisk is an interesting company with origins similar to Visa and Nasdaq. My only rub is that insurance is not a growth industry, but still, it is bound to be a great business and the stock has indeed done great. So has Nasdaq.]

Take a business like, the world’s largest genealogical service. In addition to the fixed costs associated with developing and maintaining the platform, the company has built up a database of 27 billion family history records across eighty countries, much of it purchased or licensed. [AA Comment:  His point is that benefits tremendously from network effects. The more people use it, the more valuable it becomes.]

Chapter 3: It Takes a Village: The Sources of Digital Competitive Advantage

More disturbing, as discussed in greater detail in chapter 13, is the fact that programmatic advertising allows marketers to reach New York Times readers on other websites. This reality is reflected in the dramatic reduction in absolute advertising revenue in both print and digital products and the continuous declines in advertising rates (CPMs or cost per mille, or a thousand impressions). [AA Comment: Google used to talk about programmatic advertising a couple of years ago, but they stopped doing that after it boomed. Can’t see why one day, all advertising won’t be programmatic advertising.  All that means is that a computer program is placing the adds instead of an agent on the phone, like it used to be.]

Netflix, widely viewed as a leader in customer engagement and retention, still has churn of around 3 percent each month or 36 percent annually. [AA Comment: More than a third churn out annually is quite high – at least in comparison to TurboTax, Bloomberg, Costar, or Visa. Churn is the most important metric in the economic value of a subscriber base. It’s the same with asset management and pest control.  There’s demand-side churn (customers leaving) and there is supply-side churn (employees leaving). For a service business where human interaction is critical (think Orkin Pest Control), technician churn (supply side) is as critical as customer churn (demand side). While I’m on critical metrics, Sam Walton’s favorite was “shrinkage” – the term used in retail to measure loss of inventory to theft and error and other mishaps. Low shrinkage is a reflection of strong culture.]

In the context of platform businesses, this ability to easily shift among or support the simultaneous use of multiple platforms is often given a fancy digital moniker: multi-homing. There is nothing objectionable in this term, but it is sometimes treated as an entirely novel category of phenomenon rather than simply a manifestation of weaker demand-side barriers to entry. [AA Comment: Indeed, the more aggregators there are, the less sticky a platform becomes. If most riders use Uber and Lyft interchangeable, that’s low supply-side barriers to entry.]

It is increasingly certain, given the subscriber boost from the pandemic and the Trump presidency, that the New York Times will deliver or exceed its lofty goal of 10 million subscribers in 2025. In so doing, it may finally exceed the revenue the print paper achieved a quarter century earlier in 2000. But its shareholders will likely still see a more modest bottom line. [AA Comment:  This is what I concluded about Disney when we sold it in 2015. Ten years from now, when (and if) Disney+ is most of the business, it will be a less attractive business. Knee stops short of hanging Disney in his book, which makes sense – but he leaves little doubt to the discerned reader that he believes Disney is doomed to become a less attractive company. When the future of a narrative looks worse, it’s best to sell the stock. Disney is vertically integrated in a business where aggregators rule. Its advantage in content creation has been overwhelmed by Netflix’s massive gamble. As I have been saying since 2015 – Netflix already won. I agree with Knee’s skeptical views only as one alternative angle to the problem of separating the signal from the noise. He is spotting delusions while I am trying to make dreams come true. We are bound to differ in our actions even when our minds are aligned.]

A digital New York Times in 2025 would be a smaller and less profitable business, even ignoring the cost of eliminating the printing and distribution operations.

As many significant technology businesses have few assets and little revenues, a number of competitively questionable transactions have slipped under the financial thresholds for the required government antitrust notification of the Hart-Scott-Rodino (HSR) Antitrust Improvements Act. Google had even found a loophole to avoid giving regulators a heads-up on its billion-dollar acquisition of Waze. [AA Comment: Found a loophole? Not really. Waze would have disappeared if Google didn’t buy it, because they would have to compete with Maps, and would have never been able to offer it for free for so long without Google.]

Regulators waking up to the potential dangers of big tech is welcome news. But there is a difference between identifying a problem and locating the most problematic issue and the most effective remedy. The fact that the antitrust authorities chose blocking AT&T buying Time Warner—a deal so incoherent that AT&T reversed it in less than three years—as most worthy of extensive (and failed) litigation to protect the public suggests that the chances that the government will get it right in big tech are low.

Part II: In the Land of the Giants

The five enormous companies that constitute FAANG—Facebook, Amazon, Apple, Netflix, and Google—owe their inclusion in this ubiquitous acronym to television personality James Cramer. [AA Comment: Sundar Pichai would never go on the Cramer show, but Sasan Goodarzi does. That’s about the only rub I have with Intuit.]   

Obsessing over either the precise composition or the ups and downs of FAANG distracts from a much more fundamental question: What do these businesses really have in common that investors should care about? [AA Comment: Agree. The key question is: Which is THE MOST OUTSTANDING company? He answers it throughout the book.  It’s Google, hands down.]

More fundamentally, the underlying sources of success for each of these businesses are quite diverse. Only one of the five platforms—Facebook—exhibits characteristics broadly consistent with the narrative of the Platform Delusion. [AA Comment: He states that Facebook is a victim of the delusion. In the same chapter quotes Freud! I go with Freud in predicting that the Facebook delusion is only growing, and that Zuck is not standing still. That said, why not just own more Google?]

Much is made in the press of Amazon spending more on R&D than any other company in the world. But as a percentage of overall costs, this still represents a surprisingly low number. And none of this takes into account that by reporting “Technology and Content” costs, rather than R&D like its peers, both the absolute and relative levels of spending are undoubtedly overstated when making comparisons. [AA Comment: Didn’t quite understand his point. Who overstates against what peers? Cost has never been a big part of the Amazon narrative and it still isn’t.]

Facebook spends far less in absolute terms on R&D than any of the FAANG companies other than Netflix. But Facebook and Google are the clear outliers when it comes to relative R&D spending. Both consistently dedicate north of 20 percent—and in the case of Facebook, approaching and in some years north of 30 percent—of its total costs to R&D. [AA Comment: Edwards’ R&D runs at 18% of sales. About half of it is spent on trials.]

even Google is not only far behind Baidu in China (where Google is a distant fourth or fifth) but also behind Yandex in Russia, and it has meaningful competitors in South Korea (Naver) and Japan (Yahoo). [AA Comment: Google is probably behind North Korea too. So what? And for how long?]

Although product search is a small part of overall search, given the psychic proximity of this subset of searches to spending money, it is among the most valuable. [AA Comment:  Google product search is much better than Amazon. More choices. Better organized. As the ultimate meta search engine, Google is the aggregator of aggregators.]

LinkedIn dominates professional networks and Facebook’s belated job application feature is unlikely to change that.

Chapter 4: Facebook: The Ultimate Network [FACEBOOK]

The power amassed by Facebook, the world’s largest social network, is Exhibit A in support of the Platform Delusion—Facebook is the ultimate network effects driven platform that quickly took over the global market. Facebook is a purely digital creature, something for which the analog world offers no real counterpart. [AA Comment:  It’s most attractive attribute is that it is the purest platform play among big tech. But that can also become a weakness if and when the winds change.  Facebook is all about Direct Targeting (DT).  They do it better than anybody else, including SNAP and TWTR and PINS.  They convert more ads into sales and generate more revenues per user than anybody else. They offer the advertisers the highest ROI.  As Apple pulls cross-App data sharing and Google pulls cookies, it will become harder to get access to good direct targeting engines.  This will only make FB more supreme in direct marketing. It is unlikely though, as some analysts fear, that advertisers and brand builders, will deploy less budget towards Direct Targeting. After product search, it should continue to have the highest ROI, because nothing can beat knowing your audience well.]  

Every new user is a potential new connection for existing users, instantly improving the product with no incremental effort by the company. [AA Comment: This is the essence of network effects. It’s why platforms where creators contribute content, such as YouTube and Instagram, are precious.]

The rise and fall of MySpace—which briefly in 2006 overtook Google as the most visited site in the United States—has been well documented, but years before that there was And who remembers Google’s Orkut, which predated both MySpace and Facebook and owned the Brazilian social market—until it didn’t.

where the technology is viewed as still changing, consumers will be reticent to become too attached to any product or platform, particularly any one that requires a significant financial or emotional investment.

There is a good argument that Facebook was the beneficiary of propitious timing.

But if someone were offering a demonstrably better deal—free move tickets or iTunes, anyone?—how hard would it be to convince your most important user groups to join you on a new platform? [AA Comment: Making them shift is easier than keeping them engaged. Just because someone starts using another platform for its free benefits doesn’t mean they stop using Facebook.]

When it comes to customer captivity, Facebook is the ultimate Hotel California. [AA Comment:  Yep. They make it really hard, but they have nothing on me because I have never opened an account. Not being a member often restricts my ability to watch a cool video. But it is what it is. Sometimes access is available only through Facebook, but it is almost always just a search away on Google.]

With few exceptions—we can forgive Zuckerberg the Oculus VR acquisition—every major organic and inorganic investment has been directed toward enhancing or protecting the core social network franchise. [AA Comment: Forgive him for what? Zuck has been excited about virtual reality for a while.]

Facebook’s emphasis on efficiency as well as focus, however, has proven an additional tool to both protect and fully leverage the demand and supply advantages that are needed to buttress scale. [AA Comment: Focus is truly key. But Facebook is purportedly very focused on current profits, despite what Zuck says on calls.  Google is much calmer about monetization. Even PayPal is calmer. A deferral of monetization and focus on engagement, can actually make it more likely that a platform stock is underpriced. The poster child of this is Amazon.  But Google is also an example. To be fair with Facebook, WhatsApp is too.]

Facebook has long been famous for an aggressive culture of continuous improvement—which increases satisfaction and ingrains habit while making it more difficult to successfully search out a comparable alternative.

…review of the demise of Friendster shows that after a certain point in social networks, incremental participants—particularly if they are trolls, pedophiles, catfishers, scam artists, or hostile governments—can actually detract meaningfully from value.

That the government has actually filed its long-anticipated suit seeking to break up Facebook is unlikely to have a meaningful impact on the company. Most obviously, the complaint will take years to resolve and is difficult to prove and even harder to implement.

An empire-building CEO ranting to subordinates online may not be attractive, but it is not clear that it calls for government intervention. Even if one day Facebook is forced to divest the two acquired companies, it may be an “existential threat” to Zuckerberg’s social standing in Silicon Valley, but the shareholders will do just fine. [AA Comment: Agree that Facebook will be fine. Disagree his standing will diminish in Silicon Valley. People have disliked Zuck for years.]

Earlier we noted the irony of the fact that it was trust that allowed Facebook to permanently overtake the many social networks that preceded it. Another irony, not widely appreciated, is that there is strong evidence that Facebook’s enormous scale has enabled it to combat “fake news” and other subversive forces on the internet far more effectively than its smaller peers.

The good news for Facebook, or whatever alternative “social mechanic” succeeds them, is that contrary to the simplistic conceit of the Platform Delusion, the business does not rely simply on the flywheel of network effects. If it did, recent events would have ensured a swift exodus of users to any number of competing social platforms.

focus on serving already established networks and continuously investing in tools to demonstrate the value of the platform ensured a significantly stronger level of customer captivity once scale was reached.

Chapter 5: Amazon: Can You Have Too Much of a Good Thing?

It would be almost a decade after Amazon’s founding in 1994 before it launched the “marketplace” business that does indeed benefit from network effects. [AA Comment: So what? It took T. Rowe 13 years to introduce his first mutual fund.]

But how insurmountable is this scale advantage really? … Amazon’s continuous upping of the fixed-cost table stakes in online retail is a sensible strategy to protect scale advantage. [AA Comment: Scale advantage is like late-night alcohol consumption. It starts weighing after it stops growing.]

Using Amazon as the poster child, as many do, for the case of a company for which “almost every human interaction is removed from the actual critical path in service delivery” is a stretch in light of the company’s one-million-plus employees. [AA Comment: I have often contrasted Amazon’s offerings to those of Google. One requires tremendous work to deliver the incremental dollar.  The other requires none. As  Larry Hite states in his book, The Rule (2019), the best businesses are those that make money while you sleep.]

It is the local density, not the overall size of operation, that overwhelmingly drives the economics. [AA Comment: Any good transportation analyst understands this, but it also applies to cable networks, auto parts, and pest control.]

The fact that Amazon profit margins fell in its core North American market in 2020 (as they had as well in 2019) even as the pandemic boosted sales by almost 40 percent is suggestive of the modest nature of Amazon’s scale benefits in retail.

Although Prime has been able to successfully implement membership fee increases along the way, these reflect little more than inflation rather than the full extent of incremental value the service provides.

The fact that Amazon feels a need to give away a service that is not only costly to provide but also exhibits notoriously high customer churn reflects how thin Amazon’s customer captivity is even after all the hard work.

An obscure antitrust law designed originally to protect small retailers from chain stores using their clout to impose restrictions or secure preferential terms with suppliers and manufacturers would appear to have the greatest potential application to Amazon. But that law, the Robinson-Patman Act of 1936, has been gutted by the courts and fallen into disuse by the regulators.

Why would Amazon need to buy—only to shut it down six years later—if the advantages of its broad platform were as deep as suggested? And how could the founder of Diapers then go on to build an alternative “platform” in just a year that was compelling enough to attract a $3 billion bid from Walmart to use as the engine of its competing e-commerce business?

One wonders whether the only thing that is stopping him [Bezos] from now buying Wayfair or Chewy—the online retail leaders in furniture and pet products, respectively—is the same thing that stopped Mark Zuckerberg from buying Houseparty or TikTok: fear of government intervention. [AA Comment: When I first read The Everything Store (2013), my biggest takeaway was how ruthless Bezos was at undercutting his acquisition targets before buying them. I believe we could see regulators questing that one day.]

In The Curse of the Mogul, my coauthors and I demonstrated that there was indeed a significant correlation between revenue growth and value creation among the largest media conglomerates over almost a quarter century. Unfortunately, that correlation was decidedly negative. The conglomerates had achieved growth largely through overpriced acquisitions and foolish internal projects. [AA Comment: Think Disney’s latest massive deal ahead of Iger’s exit and his victory lap auto-biography.]

Before turning to an examination of each of these in turn, let’s start with an obvious but highly pertinent observation. Over its history of growth, Amazon, like every other company, added the products and geographies with the greatest opportunity first, moving down the list to less-obvious opportunities. [AA Comment: Law of diminishing returns.  It implies that each incremental initiative is less attractive.  People have said this about ORLY’S and AZO’s store expansion program for decades. In reality (i.e. in practice), it does not have to work that way. I think Knee’s conclusion overlooks how the growth of the network creates new attractive growth options and degrees of freedom, that were not available before when the network (or platform), was smaller. This phenomenon is seen with Ansys, a simulation software roll-up.  The more verticals they explore, the more new verticals become available. Some of those they buy and others they pursue organically.  The stock has done great over the decades and the growth prospects look even more attractive today than they did 10 years ago.  Amphenol is another example of this.  When they had just connectors, there was less opportunity to bid on integrated solutions.  Now that they are deep into sensors as well, nobody can touch their offerings.  In their latest call on October 27th, Amphenol’s CEO said about their penetration of the auto electronics vertical, where electric cars are a big driver of growth: “Our strategy is not to take share from other companies. It is to make things happen in the car.”]

The online home category leader Zillow took a hit to its stock a few years ago when Amazon added a web page hinting at expansion into real estate referrals. The page soon disappeared, Zillow has continued to soar, and Amazon is mostly limited to selling small prefabricated homes—with free shipping—online[AA Comment: Interesting he mentions Zillow and Bloomberg, but does not mention Costar.  Despite being a software and platform rollup and growing top line by 25% annually over the last two decades (half through M&A), a seasoned investment banker in the space never mentions it. That is interesting evidence that the CSGP is under-appreciated and still largely undiscovered.]

The hugely disappointing performance of since it spun off as an independent company several years ago is reflective of this structural sectoral infirmity—as is the inability of the company to find a willing buyer even after it put itself up for sale. [AA Comment: was bad, but not hugely disappointing. Like Mirant and Delphi, it was spun off, with the assistance of smart bankers, to die.]

If every number one player in a market could easily become the number one player in every other market, all the markets would have a lot of number one players!

[Amazon] has taken considerable share from off-line retailers. Internationally, it faces not just these corresponding off-line retailers, but scale online players as well.

If India is Amazon’s most promising international opportunity—it certainly will be its most costly—investors can be forgiven if they don’t place much stock on the odds of the company achieving a higher return on investment than it did in China. [AA Comment: Investors can be forgiven? Just say it: You think India and China won’t work for Amazon.]

Even if Amazon lacks strong entry barriers in its original core commerce markets, it is still a great company. Whether it is worth its current valuation, however, is another question. [AA Comment: Hasn’t it always been? I’d say less so today than 10 years ago when people didn’t appreciate the strategy, nor the value of AWS.]

Today, although more goods are sold through the Marketplace than directly by Amazon, it still represents only a tiny portion of overall revenues because rather than the full price of the products sold, the company only receives a commission. [AA Comment:  They could probably charge higher commissions and get away with it. For every merchant that leaves Amazon, many more enter. This is because merchants need sales, above all, and Amazon has plenty.]

On its face, the financial wisdom of many of the additional benefits subsequently provided to Prime members, notably Prime Video, is highly questionable. Similarly, the return on investment of the move to same-day delivery, quickly mimicked by multiple competitors, seems unacceptably low.

Amazon’s formidable franchise has emerged from a combination of “relentlessness and ruthlessness” on the one hand and “a rope of many small advantages” on the other. Its aura of invincibility, however, is not justified and the potency of this mixture of attributes varies widely across markets. The overall return on investment of its future e-commerce growth trajectory is likely to remain modest. [AA Comment:  I wouldn’t be so negative.]

Chapter 6: Apple: What’s at the Core?

After Jobs’s triumphant 1997 return to the company, Ellison was the first new board member selected once the old guard was shown the door. Ellison has no doubt about what distinguishes Apple: “Steve created the only lifestyle brand in the tech industry.”

The fact that every few years brings the introduction of new luxury car brands suggests that strong incumbent brands do not represent a significant barrier to entry.

The introduction of the iPhone and iPad transformed the economics of Apple not simply because they were both wildly successful products. Rather, for the first time, the rapid adoption in both instances allowed Apple to benefit from deep network effects, which had previously eluded the company.

The real revolution began the next July when Apple introduced the much faster iPhone 3G at a $200 price point—and the App Store. At launch, the App Store carried around 500 applications, many from approved outside developers who had happily agreed to a 70-30 revenue split. Despite being “widely hailed for its beauty and functionality,”

A year after the App Store appeared, it would have 50,000 apps, which had been collectively downloaded more than 1 billion times. Ten years later the store would have over 2 million apps and 20 million registered developers. The store generated more than $100 billion of very high margin revenues over that first decade.

This led many to anticipate a replay of the earlier war between closed and open systems that had led to the resounding defeat of Apple by Microsoft and the IBM clones in the 1980s. And sure enough, smartphones powered by Android overtook iPhones in the US by 2010.

What has changed in recent years, however, is the extent to which loyalty to the Android OS has come to match or exceed this. Both operating systems retain around 90 percent of users within their respective ecosystems when they buy a new phone.

Yet ultimately Jobs knew it was all about the product. So, when he confided in Walter Isaacson that “Tim’s not a product person per se,” one wonders what Jobs really thought about the company’s long-term prospects.

…or relatively narrow niches where even great success would make a modest contribution (e.g., AirPods). Nothing, in short, is anticipated to move the needle.

Apple began to make it clear where it wanted investors to focus—and where not—in 2018. To the anger of research analysts who had long tracked the company, Apple announced in November 2018 that it would no longer report iPhone unit sales. The following quarter, Apple for the first time began reporting the relative profitability of the product and services businesses, which highlighted both how much more inherently profitable the services business was and how much faster it was growing. [AA Comment:  Not that I would dare call Apple a short, but carving out a faster growing, yet much smaller, piece of the business is how bad narratives get spun (think GM, INGR, TAP).  I can give many examples of this mental accounting practice being put to work, and most ended badly.]

Given the relative growth and profitability of services, they could account for a majority of Apple’s revenues by 2030 if current trends persist. This shift should in theory strengthen the overall Apple franchise. [AA Comment:  Ten years is too long of a time away in Tech land.]

Although it is not possible to identify precisely the extent to which the problem was the product itself or the limitations of the brand, over the years Apple has disappointed in watches, televisions, video game consoles, and more. [AA Comment: Not everyone believes that, but ok.]

As we move to the realm of ubiquitous low-priced streaming video services, the value of the Apple luxury brand is even more tenuous. Here the core value proposition is not technological or design pizazz but rather the provision of compelling proprietary entertainment content.

I subscribe to HBO to get Game of Thrones, not the other way around. [AA Comment:  Great one liner!]

Chapter 7: Netflix: Content Was Never King and Still Isn’t

“Content is king” is generally treated as a self-evident notion, universally embraced by professionals and the public alike. The origins of the phrase are alternatively attributed to media mogul Sumner Redstone and tech icon Bill Gates (neither actually coined it).

In his autobiography, Sumner Redstone, who had controlled ViacomCBS and consequently the TV network CBS and film studio Paramount Pictures, traces his epiphany about the supreme importance of content to his early days in the movie exhibition business. “You can have the most beautiful theater in the world,” Redstone realized, “but if you don’t have a hot picture, forget it.” This led Redstone to the broader point regarding entertainment viewing: “They watch what’s on it, not what it’s on!”

What modest profit the largest content players do eke out has typically come from the ability to monetize ancillary businesses in marketing and distribution that do scale. [AA Comment:  This is what Disney has done, and continues to do, best.]

Over the years, there have been plenty of arguments—with expensive consequences for a long list of short sellers from Whitney Tilson in 2010 to Andrew Left in 2019—that Netflix should not be as successful and as highly valued as it is.

In fact, the dirty little secret of the media industry is that content aggregators, not content creators, are the overwhelming source of value creation.

Although much of the public discourse surrounding Disney’s $71 billion purchase of 21st Century Fox earlier in 2019 involved the excitement over the uniting of Fox studio’s X-Men and Fantastic Four franchises with the rest of Disney’s Marvel multiverse, most of the profit of the business being purchased came from elsewhere—both the regional sports networks that Disney would be forced to divest as well as the collection of domestic and international cable networks that were kept contributed far more than the filmed entertainment division.

The structural superiority of the content aggregation business to the content creation business should not come as a surprise. The economic structure of the media business is not fundamentally different from that of business in general.

Aggregation, on the other hand, by its nature requires a large fixed-cost infrastructure to collect, manage, market, and redistribute content. This is why a cable channel with 20 million subscribers loses money but an identical one with 100 million subscribers might generate 50 percent margins.

Without the fixed-cost requirements associated with producing and distributing CDs and managing racks at Tower Records, the barriers to entry into music are not what they used to be. The detriment of increased competition simply outweighs the benefit to established businesses of lower fixed costs.

Netflix’s ability to spread the fixed costs of content, marketing, and technology across a subscriber base vastly larger than any other competitor’s is continually reinforced by superior customer service, a powerful recommendation engine, and a great, habit-forming product.

Only in the media industry, however, would it seem a paradox that owning the exclusive broadband pipe into the home at a time of exploding usage makes for a good business. Relying on dumb pipes instead of expensive content or talent is always the smart bet.

It is hard to envision a happy ending for shareholders given these continuing trends, notwithstanding the stock outperformance of not just Netflix but Disney in 2020. [AA Comment:  He seems surprised that Disney outperformed in 2020.  It hasn’t been doing that great in 2021.]

In 2018, Hulu had 25 million domestic subscribers, a little under half of Netflix’s 58 million US subscribers at the time. Just a couple of years earlier, a respected Wall Street research analyst argued that Hulu’s value was $25 billion based on a much more modest subscriber prediction for 2018. But the analyst assumed 2018 profitability. In fact, losses had accelerated along with the subscriber count, far exceeding $1 billion.

What is known is that, based on some internal Amazon documents obtained by Reuters in 2018, the number of Prime members that actually watched any video was less than a third—today, not much more than the 28 million who actually pay for a Hulu subscription.

Netflix’s decision to begin financing original films in 2015 made sense given the disproportionate share of viewing that films had always represented in Pay TV. That said, film production is a very different undertaking than producing television series.

The moral here is not that Netflix shouldn’t invest hugely in original production. Indeed, in the face of the competitive onslaught it makes absolute sense to press its relative scale to heighten the fixed-cost price of entry.

In Platform Revolution: How Networked Markets Are Transforming the Economy and How to Make Them Work for You, three consultants and academics similarly argued that Netflix has strong network effects.

It is worth noting that Netflix itself has largely eschewed claims of network effects. Reed Hastings contrasts companies like Netflix that have “normal scale economies” with “those rare businesses like LinkedIn and Facebook where there are network effects.” It was not for want of trying.

Hastings ultimately described his futile quest for network effects as a “competitive fantasy.”

As Netflix only somewhat hyperbolically likes to describe it, there are as many customized “different versions of Netflix” as there are subscribers.

This is especially true in light of the company’s decision in 2010 to stop reporting customer churn, even in the face of SEC resistance, based on the obviously specious justification that it is inadequate as a “reliable measure of business performance.”

Estimates of Netflix annual churn mostly have ranged from 20 percent to over 35 percent.

…according to Carr, “Netflix was able to find a Venn diagram intersection that suggested buying the series would be a very good bet.” This narrative is so ridiculous on its face that one would not feel a need to rebut it were various versions of this story not repeated so relentlessly. [AA Comment:  That’s an aggressive statement.]

Soon after the triumph of House of Cards, Netflix committed to a series more than twice as expensive—Marco Polo. Marco Polo was one of the first Netflix series to be canceled.

…the Nielsen ratings service claims that by 2024 it will have developed an entirely new ratings metric that incorporates digital and traditional viewing. [AA Comment: Nielsen said that 5 years ago. It is one of the biggest losers in the space.]

Birdbox and Murder Mystery are in a class of mid-budget films that are neither sequel nor spinoff that have stopped being economic for theatrical distribution—even when topped off with a big-name star (or two).85 Netflix has demonstrated that there is still a demand for some of these at least at home. What big data has not done, and will not do, is provide a template for how to make them well.

Hastings’s 2020 book with Professor Erin Meyer on the company’s culture and management philosophy, No Rules Rules: Netflix and the Culture of Reinvention, contains a number of anecdotes about programming decisions at the company. What is most notable about these descriptions is just how small a role data appears to play in practice. [AA Comment:  That is true.  I loved this book, but did not take note of this because I read it with a different purpose. The book about TikTok, on the other hand, was all about data.]

As of 2020, although most of Netflix’s new content is original production, the vast majority of what is watched remains licensed.

the winner in most markets is someone who let others undertake what is effectively free R&D for them and only invests big once greater visibility emerges as to the shape of demand and technology requirements. was the original online employment classified site and one of the earliest successful internet IPOs in 1996. After going public at $7, the stock reached a peak of $91 in 2000 at a valuation approaching $10 billion. It was bought by a global staffing company for a few hundred million dollars in 2016, having long been marginalized by direct broad-based and niche competitors like CareerBuilder and as well as entirely new competitor categories like LinkedIn and Indeed. [AA Comment:  The firm I used to work at then got killed in this name.]

In SVOD, Disney appears to have committed itself to this strategy and has the assets required to achieve scale, even if the financial returns realized on the road to getting there seem bleak. None of the other emerging competitors seems to have the combination of skills, resources, or commitment required to become a long-term global Netflix competitor. [AA Comment: Many people conclude Disney will succeed, but what does success really mean? I doubt Disney’s business will be more profitable in the future than it was five years ago, no matter how much mental accounting they do.]

While it was strategically sensible for Netflix, the business of taking creative risk has always yielded paltry long-term financial returns. [AA Comment:  You can say the same about Amazon and Tesla.]

Google’s longtime chief economist, Hal Varian, wrote a strange blog post a dozen years ago arguing that a single attribute explains the “secret sauce” behind its remarkable results: learning.

This relative success with the federal regulators was achieved despite the fact that Google represents the most impregnable competitive fortress among its FAANG brethren.

Chapter 8: Google: Letter-Perfect Alphabet

Like Netflix, Google started life as a pure aggregator—as its corporate mission “to organize the world’s information” makes clear. Unlike Netflix, however, the depth of Google’s structural advantages ensured that it never needed to go into the content creation business in any serious way.

If Microsoft couldn’t make a dent with Bing after over a decade, it is small wonder that DuckDuckGo has fared no better. Those few who can still claim relevance in the market are restricted to a protected geography (Baidu in China or Yandex in Russia) or niche search use case (Amazon in product search).

…there is little question that Google’s greater familiarity with prior search behavior drives a scale advantage on the demand side by facilitating the effective customization of the selection and presentation of search results for individual users. … So to the extent that there is a direct network effect on the user side of search, it is overwhelmingly driven by the number of one’s own prior searches rather than the number of other searchers.

AdExchange, Google’s real-time bidding exchange for premium publishers and big brand advertisers, benefits from the same network effect dynamic.

Notably, AdSense and AdExchange—collectively reported as Google Network Members’ properties—Google’s businesses that benefit most strongly from network effects, are a fraction of the size of the core search business that benefits predominantly from supply-side scale. The advertising revenue from this segment is less than a fifth of the revenue from Google’s owned search properties, including, YouTube, Gmail, and Maps. And that proportion has been falling for years.

Customer loyalty, from both searchers and advertisers, to Google’s search engine is a critical factor reinforcing the benefits of scale. Users become more effective at using particular search programs with experience, and that experience makes the incumbent search engine more effective at delivering relevant results.

…the Justice Department’s 2020 antitrust suit ultimately decided to target an entirely different aspect of the business for the time being. Even if the federal government decides to revisit this topic later, or if the state lawsuit that focuses on it is successful, it will have little impact on the overall customer stickiness of Google’s franchise or of its ability to invest more than anyone else to make the experience even better.

The track record of Google, however, suggests that search represents a use case in which the integrated advantages of data-driven learning and technology are persistent and continue to grow without topping out.

This kind of virtuous cycle between Google’s learning-enhanced proprietary technology and network effect supported customer captivity among both users and advertisers on the one hand and traditional cost–based economies of scale in R&D and other areas suggest that its remarkable economic performance is likely to endure.

Until the restructuring, Schmidt conceded, “a disproportionate part of the day would be spent on moonshots.”

Many questions remain with respect to the future of Alphabet. The Economist has expressed skepticism over whether the increasingly corporate management team has the vision to transition the famously engineering-centric culture through middle age.

Although with new leadership it has accelerated growth more recently, that Google Cloud remains a distant third is reflective of the deep challenges in building a sales culture from a standing start.

As early as 2014, the company realized that an “over-glamorization” of moon shots at the expense of “methodical, relentless, persistent pursuit” of opportunities closer to home had been costly to the company.

Apart from Google’s clever government and public relations efforts, the federal government’s decision to tailor its challenge narrowly to Google’s commercial deals with Apple and others to serve as the default search engine is likely driven by one primary consideration: it’s a winner.

In its most recent 10-K, Google quietly removed its insistence that it would not pay a dividend “for the foreseeable future.”

Perhaps the most important lesson of a close study of Google is the most obvious one: wherever possible, at least when it comes to its core advertising franchise, avoid competing with it.

While there is value to analyzing the impact of each particular competitive advantage, this should not distract from the overarching insight that Google’s singularity stems from the unparalleled breadth and depth of its collection of mutually reinforcing competitive advantages.

Part III: In the Shadow of the Giants

Chapters 10 and 11 are dedicated to travel and tourism, which contribute something approaching $10 trillion to world GDP.1 Although it is not, as sometimes falsely claimed, the world’s largest industry,2 it does display a number of other characteristics that justify two full chapters. The sharing economy is responsible for a disproportionate number of the highest-profile digital IPOs of recent years. The companies that fall under its rubric represent quintessential platform companies that create value by connecting holders of excess capacity with potential users of it. Chapter 12 documents the huge variations in the attractiveness of some of these businesses, despite the obvious similarities, particularly between the two largest—Airbnb and Uber.

Chapter 9:  E-Commerce: If Amazon is the Everything Store, What’s Left to Sell?

Consumer proclivity to shop online varies dramatically by product and Amazon’s relative success online varies as starkly by category. It is notable that two of the relatively small number of large-cap consumer internet companies are e-commerce retailers in categories that one would have expected Amazon to dominate if the Platform Delusion were true.

Amazon does not and is not likely to play a leading role. As noted, while the transaction volumes in these markets have continued to swell, the economics have deteriorated. These platforms are squarely in line with the long-term underperformance of e-commerce platforms generally.

During the three years leading to 2018, the percentage of e-commerce activity overall represented by marketplace models grew by over 50 percent—jumping from only 30 percent in 2015 to approaching half of all online retail transaction value. As of 2019, marketplaces represented 57 percent of the global e-commerce market.

Research firm BTIG has noted that investors in independent marketplaces are “perpetually petrified of competition from Amazon, Facebook, and Google.” After reviewing the data, however, they concluded that “investor concerns are largely unwarranted.” of the most serious charges against Amazon that is subject to regulatory scrutiny is that it actually used data from its own sellers to launch competing products—hardly an indication of an unwillingness to offend vendors on its platform. [AA Comment:  This is my biggest rub with Amazon.  It actually uses customer data to compete against the customer.  That’s not a win-win.]

At the time of its IPO a decade later, almost 90 percent of Etsy’s traffic was still secured organically rather than through search or paid channels.

Sotheby’s, arguably the leading off-line brand for expensive antiques, had lost millions trying to build an online platform—including through failed joint ventures with Amazon21 and later with eBay.

when Etsy raised its commission from 3.5 percent to 5 percent, the stock soared despite predictable seller grousing based on investors’ correct prediction that it would have no negative impact on the number of merchants.

The playbook involves copying those attributes Amazon has now established as the price of entry in e-commerce (e.g., service and shipping) but overlaying a deeply customized variation for a product category and community that is challenging for Amazon to effectively replicate. The ability of Wayfair to challenge Amazon as the leader in online furniture sales more broadly is further evidence of this.

Etsy confronted a much more direct attack with the launch of Amazon Handmade in 2015, just months after its IPO. Its shares struggled for years as it took incremental hits with every new announcement from Amazon—for instance, establishing the “Amazon Handmade Gift Shop”44 and the availability of Handmade product for immediate Prime Now delivery in certain cities.45 Performance only turned around when investors noticed that Etsy’s organic growth was actually accelerating in the face of the Amazon onslaught.46 This suggested the surprising possibility that Amazon’s marketing efforts served mostly to draw attention to the category, benefiting Etsy as the category leader[AA Comment:  I call this the Robert Wilson Theory. It happened to Intuit when Microsoft came after its accounting software, Quicken, and with Edwards when Medtronic came after its heart valves.]

If Etsy and 1stDibs demonstrate how the combination of strong network effects bolstered by product complexity and fragmented buyers and sellers can yield remarkably resilient digital marketplace franchises, their example does not demonstrate that these qualities necessarily imply invincibility

Despite the intuitive appeal of selling auto parts online, as a category overall, it is one of the least e-commerce enabled.

Amazon launched a vehicle portal in 2016 that offers increasingly complex tools to identify the right part for specific models along with reviews and advice. This level of functionality has become table stakes for anyone looking to compete in the sector.

The four giant retailers would make a huge mistake if they simply rested on their structural advantages. With smart investments that leverage their unique assets, they should be able to maintain long-term growth and above-average financial returns—even as pure digital continues to capture some more market share.

Some estimates suggest that Amazon and eBay divide a $10 billion marketplace business 60-40, and in certain product and customer categories, eBay probably remains the leader.

As Amazon faces attack from broad-based off-line incumbents like Walmart and Target, who are pursuing their own hybrid strategies, specialization and complexity will continue to provide meaningful protection for incumbents and opportunity for innovative insurgents.

Chapter 10: Fly Me to the Moon: Who Makes Money When Air Travel Goes Digital?

In the larger consumer services sectors, Amazon’s track record is even weaker. Its financial services initiatives in payments, lending, and credit cards have at best failed to make a dent and often, as in the case of Amazon Wallet, have completely collapsed.

In addition to the substantial off-line investment required to establish and maintain a consumer brand, all of these businesses needed to dedicate substantial resources to secure a favorable relative position at the gateway to most consumer journeys: Google.

…when Google bought its own metasearch company, ITA Software, in 2011 to power its flight-search tools, this accelerated the increasing importance of the metasearch channel.

Unlike OTAs, GDSs’ scale advantages are strongly reinforced by strong customer captivity on both sides of the market. Remember that GDSs started life as among the first enterprise software companies on record, an industry characterized by long-term contracts and high switching costs.

Amadeus, the overall leader in this segment as well as the industry, now earns over 40 percent of revenues from IT solutions.

Like the airlines’ establishment of the GDSs, the original credit card companies were created and owned by the banks themselves. Although these companies did well after they achieved their independence, many expected that Visa and Mastercard would be disintermediated with the birth of the online payments industry. In the early days of first mover PayPal, the company tried to do just that by incentivizing customers to use their bank account information rather than credit card numbers. But in 2016, PayPal eventually realized that trying to go around rather than leveraging the incumbent credit card networks would dramatically slow its own growth—and most significantly was creating opportunities for fast followers like Apple Pay and Android Pay.38 The resulting growth of PayPal39 and of the broader online payments sector that has emerged has only served to accelerate the value appreciation of the credit card networks. The shares of both Visa and Mastercard appreciated more than ten times over just the last decade. And while the new online payments sector has exploded, with not just PayPal but dozens of internet unicorns created, none of these new platforms come close to the value of either of these two over fifty-year-old incumbents. Indeed, that entire industry is probably smaller than the combined value of Visa and Mastercard.

Chapter 11: “To Travel is to Live!” How Priceline Became Worth $100 Billion

What is now Booking Holdings started life in 1997 as Priceline .com, the “Name Your Own Price” business made famous by commercials starring William Shatner. Although the company only changed its name in 2018, almost two decades after it went public, the reliance on the original sketchy business model had been minor for over a decade and the connections to the sketchy founder completely severed for even longer.

By the time Walker left the company and resigned from the board at the end of the year, Priceline’s shares had sunk to little over $1.13. He sold most of his remaining stake in the company over the next summer, largely severing his ties. … Icahn’s continuing enmity for Walker led him to be one of the few winners from Priceline’s stock collapse; he shorted it all the way down.

Shortly after 9/11, General Atlantic pulled the plug on funding. The failure of these organic efforts laid the groundwork for the transactions that would transform the company.

But Booking, although it has diversified into metasearch and some mostly related software businesses, is still overwhelmingly an OTA. The ability of this single company to dwarf the size not only of the entire GDS industry but all the other public companies in the OTA sector combined is attributable primarily to one factor: the difference between hotels and airlines.

When it comes to product complexity, hotels are a different matter altogether. Visiting New York for a romantic getaway? There may be only a few airlines flying your route, but once you arrive, there are almost seven hundred different hotels, of which almost five hundred are in Manhattan. And these hotels have more than a hundred thousand rooms, only some with a view.

For example, there are many drivers of the respective fates of the world’s two largest financial data providers, Reuters (later Thomson Reuters and now Refinitiv) and Bloomberg. But the single most important factor that enabled Bloomberg to overtake Reuters even with a century-long head start was that Bloomberg targeted fixed-income markets while Reuters’s historic core franchise was foreign exchange. Not only do the number of outstanding bonds dwarf the number of stocks, but the number of financially relevant terms of each—from call dates and premiums to indentures and change of control terms—is vast. Notably, this very complexity lends itself to the development of sticky software and analytic tools to track, manage, and compare the various securities. [AA Comment: It’s one of the reasons Moody’s and S&P add so much value.]

The Hotel Association of the City of New York represents more hotels than the International Air Transportation Association represents airlines globally.

Where airlines are able to enforce a roughly flat $5 per flight fee, with the GDSs able to retain at least half from all but the largest travel agents, hotels typically pay travel agent commissions of between 15–30 percent with the GDSs receiving a tiny portion for the use of their distribution infrastructure.

The bad news for the OTAs is that the largest hotel chains keep commissions closer to 10 percent.

The good news is that the top five hotel brand companies only control around half the US’s hotel rooms (the top four airlines control two thirds of the market) and far less internationally.

Best Western CEO David Kong conceded, “We basically repeated the same mistake again.”

From the end of 2009, the year that it permanently overtook Expedia’s equity value, until the end of 2019, just before COVID-19 disproportionately devastated all the travel-related equities, Booking Holdings shares had grown almost tenfold. That’s more than double the compound growth rate of the overall market. Meanwhile, Expedia actually lagged the market over this period.

Expedia relies more on low-margin airline bookings. Expedia’s revenues are also much more derived from the US, the market in which both the airline and hotel industries are most concentrated.

A subtler but equally notable distinction between the two businesses, however, relates to two different ways of selling hotel rooms—the merchant model and the agency model.

With the agency model, all that needs to be agreed on is the commission, but merchant deals involve negotiating a more detailed contract covering the net price and the inventory that will be made available. When trying to build scale quickly in a business characterized by network effects, speed is critical.

Booking had a significant advantage over companies like Expedia who were committed to the merchant model as the “superior” approach. Expedia had looked at both Active and Booking but passed because, as former CEO Dara Khosrowshahi conceded, “we were attached to the merchant model.”49 Specifically, the company had become used to benefiting from the attractive working capital characteristics of holding on to travelers’ money and the higher room markups available.

Booking Holdings has built a powerful franchise benefiting from both demand- and supply-side scale and reinforced by the customer captivity of those hotels who have come to rely on its marketing might and various software tools to attract and manage customers. But it turns out that even Booking, which dwarfs the entire GDS industry and all of its OTA competitors combined, is not the biggest kid on the travel industry value chain block. That moniker belongs to our old friend Google. When Google has a bull’s-eye on your business, you have every reason to be very afraid.

Within what quickly became a crowded space, TripAdvisor stood out for the intrinsic power and ingenuity of its model. Founded in 2000, by the time it went public in 2011, the site boasted over 50 million reviews and had established itself as an indispensable content destination for prospective travelers. The overwhelming traffic demonstrated the network effects of the model—travelers want the most recent relevant reviews of properties they are considering and reviewers want to share with the widest possible audience. As recently as 2018, the Guardian newspaper said, “TripAdvisor is to travel as Google is to search, as Amazon is to books, as Uber is to cabs—so dominant it is almost a monopoly.”

TripAdvisor remains a metasearch company and it relies on advertisers (of whom Booking and Expedia are by far the largest) who want access to the users planning to take a trip. And although the strength of its unique network effects driven review content helps its position in organic search results, nothing provides a long-term solution to sitting between Google and your two biggest customers, all of whom have their own competing metasearch capabilities.

The failure of TripAdvisor as a network effects driven digital platform to deliver the kind of performance predicted by the Platform Delusion highlights the critical importance of other structural attributes—here the lack of diversity of key network participants and low switching costs—in determining success.

Chapter 12: It’s Nice to Share, Sometimes: Why AirBNB Will Always be a Better Business than Uber

The basic economic problem of how to optimize asset utilization is not new. In the case of assets under shared ownership, this topic captured the attention of economists almost two hundred years ago under the rubric of the “tragedy of the commons.”

…the existence of network effects in itself tells an investor relatively little about the attractiveness of a particular business.

…even between apparently similar business models within a relatively narrow definition of sharing platforms, there is significant variability in quality.

Although Uber had consistently been valued at a multiple of Airbnb, the key market and product attributes that drive sustainable franchise value suggest that it is Airbnb that has always been the better business.

…it is far from clear that the ride-sharing market—unlike the space-sharing market—is really global. As a result, in any given market, Airbnb is likely to face the same handful of players while Uber is more likely to also confront significant local or regional champions.

Two primary attributes are responsible for the superiority of Airbnb over Uber: product/service complexity on the demand side and the fixed-cost requirements on the supply side. The former determines how many network participants are needed for a viable product and the extent to which additional network participants continue to enhance the product. The latter determines basic break-even economics and the relative financial advantage of being larger than competitors.

In ride hailing, other than price, the ability to deliver a car within three to five minutes dominates all other customer considerations.

In the short-term lodging market, by contrast, there are many more salient product characteristics and market segments.

…evidence suggests that more listings not only attract more travelers but also drive higher occupancy rates. This dynamic reinforces the value of relative network scale on the demand side for Airbnb that is far greater than for Uber.

some markets, like Brazil, competitors number in the hundreds, representing a mix of global giants like China’s Didi and low-cost services with basic homegrown apps.

by being a fast follower, Lyft historically has been able to free-ride on Uber’s investments to clear the regulatory way for the service.

…although a small minority of riders overall currently use multiple ride-hailing apps, that percentage is growing fast and varies widely by geography and demographic. Among my MBA students in New York City, it is already greater than 90 percent.

…the fact that Booking claimed in 2018 to have surpassed Airbnb in alternative accommodations listings, but generated only $2.8 billion of revenue from them that year compared to Airbnb’s $3.6 billion suggests the inherent value of specialization within the category.

Chapter 13: Mad Men, Sad Men.  Advertising and AdTech Meet the Internet

The precipitous fall in how much an advertiser will pay for a digital impression corresponds to the exponential expansion in available online advertising inventory. What’s more, now that technology allows users to be followed around the internet by advertisers, attracting a unique audience only gets you so far: programmatic advertising software can deliver the exact same users when they are on some other site at a lower price.

Online advertising on both desktop and mobile had been facing a relatively steady decline in their respective rates of growth for almost a decade by the time digital came to overtake traditional media in 2019. More concerning is the extent to which the growth has been disproportionately captured by Google and Facebook. Some analysts have tried to demonstrate that in some years, all of the net growth in digital has accrued just to these two players—or that the rest of the digital advertising universe was actually shrinking.

The slightly longer answer is that these two businesses operate in exceptional domains where technology provides the ability to leverage their relative scale to deliver continuous, uncapped improvements in the effectiveness of the advertisements that they deliver. It is not just that they both have loads of data, it is that they have data of the kind that is uniquely relevant to advertisers. Netflix, with over 200 million subscribers glued to their screens around the world, generates lots of data. But it historically hasn’t bothered with advertising not only because of the risk of alienating its customer base but also because the data about what shows and movies you like is just not that pertinent to advertisers.

Your web of social interactions and communications (and increasingly what you buy online and off-line) is the exclusive province of Facebook. This is the kind of data that machine learning and, yes, artificial intelligence can transform into increasingly accurate predictions about which ads will be most impactful to which users. So Google and Facebook dominate online advertising because they deliver an increasingly more effective product than anyone else can.

In 2020, Google announced that it intends to eliminate third-party cookies from its dominant Chrome browser by 2022. This may be good for your privacy but it is also definitely good news for Google’s business.

By January 2017, Fred Wilson of Union Square Ventures was predicting that the “adtech market will go the way of search, social and mobile as investors and entrepreneurs concede that Google and Facebook have won and everyone else has lost.”

No industry has been more impacted by the transformation of the advertising industry landscape than the giant advertising agencies. As they face their collective existential crisis, the sector has overwhelmingly decided to literally bet its future on the anticipated adtech-martech convergence.

the relative growth of online has driven and continues to drive decreases in the relative share of advertising going to branding rather than performance campaigns.

Fortune 500 companies that dominate ad holding companies’ client lists—the CPG giants remain a quarter of their revenues—themselves now represent a much smaller fraction of overall advertising.

Increasingly, brands themselves—most recently these include McDonald’s, Nike, PayPal, and Walmart—are making adtech and martech acquisitions.

This confusing cacophony of technologies seems a tailor-made opportunity for a different kind of business from ad agencies to exploit: consultants. [AA Comment:  He mentions Accenture once after this comment, but only as part of a list of consultants.]

…the consultant’s inorganic growth has not just been through technology buys but through the purchase of creative agencies as well.

Interpublic, Publicis, and Dentsu have collectively spent well over $10 billion—in the case of Publicis, almost as much as all of their previous acquisitions combined—on data-driven marketing solutions companies.

The case of Publicis is particularly instructive. In 2015, the company purchased digital marketing consultant Sapient for $3.7 billion. At the time, Publicis claimed that the combination would represent the “agency of the future” by establishing itself as “a leader at the convergence of marketing, commerce, consulting, and technology.” Less than two years later, the company would be forced to write down $1.5 billion of the purchase price after the hoped-for benefits failed to materialize.

In adtech, the opportunity to create a scale player to aggregate the fragmented inventory outside of the two massive walled gardens has been realized by the largest demand-side platform (DSP), The Trade Desk (TTD). Founded by two former Microsoft executives, TTD outflanked competitors by positioning itself as a friend to the advertising holding companies, eschewing efforts to disintermediate them by going directly to their clients. This allowed TTD to quickly gain relative scale by aggregating these massive sources of advertising demand and attracting publishing partners offering placement. The resulting network effects were reinforced by the supply-side scale benefits from continuous investment in enhanced software tools, the captivity of close customer relationships reflected in percent customer retention, and clear opportunities to leverage their transactional data to drive continuous improvement. [AA Comment:  TTD has a high revenue concentration with the agencies that are being disintermediated.]

In little more than a decade, TTD’s market value has grown to exceed $20 billion and dwarfs the dozens of competing independent DSPs.35 But for all its success and structural advantages, TTD in 2020 generated far less than $1 billion of revenue. Its platform placed only about $4 billion of digital advertising spending out of 2020 US spending in the category of over $150 billion. While this suggests a large potential untapped market, it also highlights how limited the opportunity outside of the massive walled gardens (which have their own huge competing DSPs) really is today in adtech. One related area that has driven a recent minor renaissance in adtech has been the explosion in the number of connected TVs, where Google and Facebook do not have the same lock on the market. This has spawned a sudden burst of both deal activity and significant new investments from independents like TTD and venture firms. [AA Comment: TTD COMPETES WITH Google’s DSP. As Knee pointed out before but did not repeat here, that’s a precarious position.]

One interesting success story is that of the standalone software business that remained after Acxiom sold its marketing solutions business to ad giant Interpublic for $2.3 billion in 2018. [AA Comment:  Not so sure of that. Axiom looks good inside Interpublic, but it was not that great when it was independent. It was frequently vulnerable to regulatory and competitive attacks, which is probably why it sold to an agency.  That said, Interpublic is the most successful agency for now.].

Chapter 14: Big Data and Artificial Intelligence: When They Matter and When They Don’t

A platform that has more information about similarly situated borrowers’ actual risk profile because it has serviced more of these loans should be better at pricing the risk of a given borrower. A less sophisticated platform might attract borrowers by offering better terms than their risk justifies but, once the resulting default rates become evident, lenders will flee the platforms. Over time as the relative superiority becomes clearer and relative market share follows, this data advantage should grow.

Interestingly, in one segment of the market, it appears that there is meaningful incremental value from data beyond simple credit scores. For borrowers with lower credit scores, additional data analysis can yield significant insight into the probability of repayment. Yet, ironically, LendingClub sought to distinguish itself by operating exclusively in that portion of the market where big data add no appreciable value.

Exhibit A for this view of the coming horizontal world order presented in Competing in the Age of AI is the remarkable turnaround at Microsoft engineered by CEO Satya Nadella. The company had lost half of its value from its highs of 1999 to the lows of 2009. The shares grew almost tenfold over the subsequent decade, becoming in 2019 the third company (after Apple and Amazon) to reach a trillion dollars in market capitalization. In the chapter called “Becoming an AI Company,” Professors Iansiti and Lakhani demonstrate how Nadella reoriented and refocused the company during this period.

…citing Google’s entry into the auto industry, Iansiti and Lakhani argue that traditional industry boundaries are fast disappearing and that the power of AI will drive the emergence of massive enterprises with continuously increasing, mutually reinforcing competitive advantages of “scale, scope, and learning.”21 In this new world of “unprecedented scale,” “specialized capabilities” will necessarily become “less relevant and less competitive.”

What was truly revolutionary about the SaaS model was that it allowed a single instance of the software to serve multiple clients simultaneously, a so-called multitenant architecture. And what vastly expanded the market potential was the increasing acceptance, and ultimately preference, of even global multinational businesses for SaaS applications.

One of the earliest SaaS companies was Salesforce. The company was founded in 1999 by a voluble and brilliant former Oracle employee, Marc Benioff, who had been inspired to build a company that delivered business software as securely and reliably as Amazon delivered consumer products.

…as suggested by Professors Iansiti and Lakhani, although the early SaaS applications were specialized, as the industry develops, the superiority of horizontal applications becomes increasingly clear by applying AI to centralized capabilities across vertical use cases.

Like the conventional wisdom underlying the Platform Delusion itself, the truisms of the SaaS revolution are all demonstrably false. [AA Comment:  Demonstrably false? I don’t agree.]

For automotive dealerships, to name one example, the market for dealer management systems (DMS) continues to be as dominated by the same two traditional industry leaders (CDK and Reynolds and Reynolds) today as it was in 2000. [AA Comment: CDK has been a terrible stock.]

In 2018, the leading global ERP provider, SAP, announced that it was effectively giving up any attempt to compete in the arena by agreeing to serve as a BlackLine reseller.

The BlackLine case highlights not only the continuing, perhaps even increasing, value of specialization in the age of AI but also the essential role of human judgment in deciding when applying machine learning techniques is a good use of corporate resources and when it is a costly distraction.

…often it is in the context of specialized data sets that machine learning can yield the most compelling insights.


The ability of Wayfair to thrive in the shadow of Amazon in the relatively mundane online furniture market reflects the value of specialization versus absolute size. And sometimes just not being one of the tech titans can be a source of advantage by establishing the alternative independent source, as in the case of The Trade Desk, or where trust and data security is critical, as in the case of LiveRamp.

There are very few massive markets, most notably search, that lend themselves to winner-take-most dynamics. Even social media, which in the form of Facebook has many of the same structural barriers as Google, has shown itself to be much more vulnerable to targeted geographic-, demographic-, and product-based competitive attack. [AA Comment:  In other words, just own Google.]

most justified is in product categories that essentially did not exist previously.

…the good news and the bad news is that the ambitions of the most-sought-after MBA graduates have changed dramatically. Today over 50 percent of Harvard Business School’s graduates join small, earlier stage businesses, with over 10 percent of the class of 2020 for the first time electing to do something on their own. Almost 20 percent of Stanford MBAs actually start their own businesses and fewer than 1 percent go into investment banking anymore.


September 7, 2021:  Columbia’s Knee on the Big Tech ‘platform delusion’ (2,977 views):


  • The Accidental Investment Banker: Inside the Decade that Transformed Wall Street (2006)
    This tell-all chronicles Knee’s time at Goldman Sachs and Morgan Stanley, revealing a world that rivals 24 in intrigue and drama.” Investment bankers used to be known as respectful of their clients, loyal to their firms, and chary of the financial system that allowed them to prosper. This tell-all chronicles Knee’s time at Goldman Sachs and Morgan Stanley, revealing a world that rivals 24 in intrigue and drama.” Investment bankers used to be known as respectful of their clients, loyal to their firms, and chary of the financial system that allowed them to prosper.
  • The Curse of the Mogul: What’s Wrong with the World’s Leading Media Companies (2009)
    If Rupert Murdoch and Sumner Redstone are so smart, why are their stocks long-term losers? We live in the age of big Media, with the celebrity moguls telling us that “content is king.” But for all the excitement, glamour, drama, and publicity they produce, why can’t these moguls and their companies manage to deliver better returns than you’d get from closing your eyes and throwing a dart? The Curse of the Mogul lays bare the inexcusable financial performance beneath big Media’s false veneer of power.By rigorously examining individual media businesses, the authors reveal the difference between judging a company by how many times its CEO is seen in SunValley and by whether it generates consistently superior profits. The book is packed with enough sharp-edged data to bring the most high-flying, hot-air filled mogul balloon crashing down to earth.
  • Class Clowns: How the Smartest Investors Lost Billions in Education (2016)
    The past thirty years have seen dozens of otherwise successful investors try to improve education through the application of market principles. They have funneled billions of dollars into alternative schools, online education, and textbook publishing, and they have, with surprising regularity, lost their shirts. In Class Clowns, professor and investment banker Jonathan A. Knee dissects what drives investors’ efforts to improve education and why they consistently fail. Knee takes readers inside four spectacular financial failures in education: Rupert Murdoch’s billion-dollar effort to reshape elementary education through technology; the unhappy investors—including hedge fund titan John Paulson—who lost billions in textbook publisher Houghton Mifflin; the abandonment of Knowledge Universe, Michael Milken’s twenty-year mission to revolutionize the global education industry; and a look at Chris Whittle, founder of EdisonLearning and a pioneer of large-scale transformational educational ventures, who continues to attract investment despite decades of financial and operational disappointment. Although deep belief in the curative powers of the market drove these initiatives, it was the investors’ failure to appreciate market structure that doomed them. Knee asks: What makes a good education business? By contrasting rare successes, he finds a dozen broad lessons at the heart of these cautionary case studies. Class Clowns offers an important guide for public policy makers and guardrails for future investors, as well as an intelligent exposé for activists and teachers frustrated with the repeated underperformance of these attempts to shake up education.

Recent Book Reviews

Book Reviews

Idea Hub