Reading "The Streaming Book" by @ballmatthew

Aligning on Streaming’s Business and Delivery Models

Just as consumer behaviors and preferences eventually become irrefutable, it eventually becomes possible to differentiate between "interesting” and “viable” technologies and business models. Slingbox, for example, never grew beyond its base of tech aficionados. Collecting pay-TV sent from faraway network centers, then relaying it to your mobile device, was too inelegant and bandwidth wasteful to scale. The long-term legality, too, was debatable.

For a time, TV Everywhere was imagined as pay TV future – Reed Hasting even identified it as Netflix’s second greatest threat. TV Everywhere solutions involved a household using their pay-TV credentials (i.e., for their online account management and billing) to log into the apps operated by each TV network. But like Slingbox, TV Everywhere proved too cumbersome for mass adoption and, like Slingbox, it was a neither-here-nor-there solution. While TV Everywhere aggregated billing and authentication for a pay-TV customer, users had to download standalone apps for every channel they wanted to watch (and download them not from their pay-TV provider but from a device-maker). The act of “changing the channel” involved closing one app, opening another, and navigating through its unique interface and content rights (which often differed from what was available on linear TV). Not a solution suitable for the 300MM Americans who watch TV daily.

For a time, many Hollywood executives hoped that Hulu would bring “pay TV” into the digital era. The company was founded by 20th Century Fox and NBCUniversal (Disney joined in 2009) with the purpose of aggregating TV and film content (after their linear and theatrical windows) into a single D2C platform. As all of Hulu’s content would be second window (or later), Hulu’s shareholders believed that the service would complement their existing business. And by selling D2C, rather than through pay-TV distributors or licensing to Netflix, they hoped the service would drive substantial revenues and at high margins. However, the complexities of the joint venture model and fears of pay-TV cannibalization ultimately stymied the company’s growth. For example, Viacom, originally slated to co-found Hulu, dropped out, while Time Warner also declined (it would join nine years later). And despite investing hundreds of millions into Hulu, each owner hedged their bets by developing their own standalone apps. There were also disagreements about when content should be made available to Hulu versus these standalone apps. Disney’s ABC placed its content on both Hulu and WatchABC the day after it aired on ABC, but 20th Century Fox placed its series on Fox the next day and waited a week before placing it on Hulu. The result was consumer confusion and also  tension on the corporate boardroom level, for instance over the prospect of Hulu producing original series that, heaven forfend, might compete with the pay-TV ecosystem and the networks of its owners. Furthermore, the commitments from its owners were also limited to (in a somewhat simplified sense) the last four episodes that aired on their broadcast networks (e.g., ABC, NBC, FOX but not The Disney Channel, USA, FX, etc.) and were produced by their internal studios (e.g., ABC Studios airing on ABC, not a show made by ABC Studios airing on NBC). This further exacerbated the palpable confusion and fragmentation. NBCUniversal even launched SeeSo, a D2C streaming service focused on its comedy content—meaning that both the NBC app and Hulu app would both have a smattering of NBC’s most recent episodes, but SeeSo would have a years-deep catalogue of NBC’s comedy series, such as Parks and Recreation and Saturday Night Live. Little wonder that when some executives and analysts argued that the solution to TV Everywhere’s “many apps” problem was simply to make Hulu the sole home for TV Everywhere content, the idea never took off.

The inadequacies of TV Everywhere and Hulu led to the rise of vMVPDs, which aggregated all live TV into a single app. In effect, vMVPDs were identical to traditional pay TV, such as Spectrum or Xfinity, but dispensed with the set-top box and truck roll installation and were OTT-only. For a while, some argued that vMVPDs would ensure that the pay-TV bundle would remain the dominant model of video distribution, not SVOD. The logic here was always thin, but there was a period when topline figures had the capacity to deceive. From 2015 to 2019, vMVPDs were essentially holding up the entire pay-TV ecosystem by adding 1–2.5MM subscriptions annually and averaging 150%+ compound quarterly average growth.

But the vMVPD model never addressed the core problems with traditional pay TV. For example, while vMVPDs were cheaper than traditional pay-TV service, they were typically watched 50–75% less as well. As such, they failed to correct the price/value equation of pay-TV service. Furthermore, the lower price of vMVPD service meant that most operators were gross margin–negative on the product—and operators had no clue how to solve it. In August 2018 AT&T Communications CEO John Donovan told The Information, “But I don’t think [any vMVPDs have] it figured out… And that’s why I said [that DirecTV Now is] a placeholder, because we needed to have something in that sort of mid-range price point.... What we’re trying to do is figure out how should we reshape it over time.” Four years later, and despite tens of millions of homes cutting their traditional pay-TV cord, DirecTV Now (divested by AT&T in 2021 alongside the rest of DirecTV) had lost nearly two-thirds of its subscribers, in part due to a constant string of price hikes intended to, well, stop losing money. In May 2018, Sony mentioned PlayStation Vue, its bid in the vMVPD space, just twice in a 67-page investor relations presentation; one mention came under the heading “Challenges and improvements,” stating that Vue’s “Market & future business model remains uncertain.” Seventeen months later, Sony announced it was shutting down the service, making it both the first to launch and first to go. First-mover advantages, after all, can’t fight economic gravity.

The companies that have stuck with the vMVPD business have been forced to ratchet up their price points. While this prevents ruinous losses, it also substantially reduced growth. Subscribership is growing at 9% year-over-year, down from 100–200%, or 1.2MM gross, even as the traditional MVPD system sheds 7–8MM subscribers annually (thereby putting an end to vMVPDs’ ability to sustain the MVPD system). vMVPDs are still cheaper than their substitutes, as they don’t require set-top-boxes or costly truckroll installations, but not cheap enough to make a difference. In March 2023, YouTube TV hiked its entry level package from $65 to $73 per month – more than twice its debut price of $35 in February 2017.

To this end, the primary challenge faced by vMVPDs remains much lower-cost and easier-to-use SVOD services. And with every year that passed, these substitutes generally offered more programming. Making matters worse, most of the newest and fastest-growing SVODs were operated by the very networks that supplied vMVPDs with their content. And every year, these SVOD services were becoming more important to their parents, shareholders, audiences, series, and content creators. On several occasions, parents were even placing some of their highest-potential series exclusively on these SVOD services, skipping pay TV altogether. To launch CBS All Access in 2017, CBS Corporation even made The Good Fight, the sequel to its globally acclaimed broadcast series The Good Wife, exclusive to the SVOD, thereby forcing their traditional TV fans to leave traditional TV to keep watching a series from traditional TV!

There is some future for vMVPDs. Hulu with Live TV is sustainable, if only because two-thirds of its content costs are just transfer payments to its own parent Disney/Fox. YouTube TV is believed to be gross margin–profitable because it leverages YouTube’s enormous digital ad business and scale. But with SVOD adding tens of millions of subscriptions annually, traditional pay TV losing 7–8MM (and vMVPDs regaining less than 15% of these subscribers), and the latter losing its best content to the former, it’s clear what is and isn’t the future of TV.

If vMVPDs proved to be the worst “streaming business,” UGC video platforms established themselves as by far the “best” ones, offering more revenues, more content, more profits, more users, more engagement time, more valuable “data.” Meta is estimated to generate more than $30B per year in video advertisements in the Facebook and Instagram feeds, with 0% revenue share. YouTube’s revenue is even higher, though on an unspecified portion of this revenue, it pays out 55% to creators. In 2022, TikTok generated more than $12B in revenue, with 50% revenue share for top creators. The company’s valuation is nearly $300B—more than Netflix and Disney combined. Interestingly, live broadcasting has not proven to be a good business in the West. General-purpose livestreaming platforms, such as YouNow, never took off, while live functionality on YouTube and Facebook failed to break out. Twitch is popular, with nearly 8B hours watched in the United States per year but is believed to generate less than $1B in revenue annually and remains unprofitable. In China, such services, not limited to game streaming, are quite lucrative. AnyColor, a tech company that provides avatar technologies to live streamers, is listed on the Tokyo Stock Exchange and holds a billion-dollar-plus valuation in US dollars.

An industry joke runs that there is one unequivocal winner of the non-UGC “Streaming Wars” – social media companies, such as Facebook and Twitter, that benefit from SVODs bidding endlessly against one another to advertise their (unprofitable) services to social media users (which will probably have a negative lifetime value). Another argued set of winners includes those that sell these money-losing SVODs, such as Amazon, Apple, and Roku, which facilitate access to streaming services. To start, this was through hardware that enabled “dumb TVs” to receive OTT video, then by aggregating a customer’s access/billing/search/recommendations across myriad SVODs, and then by serving as the “OS” for TV makers. Any business that sells software via software—that is, goods with no inventory or marginal costs, and no minimum guarantees—is generally compelling. Especially when they own the customer’s payment information! But the reality is a bit more complex.

In recent years, the number of subscriptions sold by these platforms has grown severalfold, as has their share of total SVOD subscriptions. In 2015, there were roughly 45MM US SVOD subscriptions, over 90% of which were billed directly (e.g., a subscriber signed-up to Netflix at, with 10% resold by a platform such as iOS, Roku, and Amazon. Four years later, the number of subscriptions is nearly 220MM, with 31% billed by platforms and only 69% direct. While this two-vector growth is important and valuable, it also reiterates one of the core problems for connected TV devices—most usage goes unmonetized. And it’s likely that this unmonetized portion far exceeds the 69% figure.

Take Netflix, as an example. It has more subscribers in the United States than any other SVOD service (roughly 65MM compared to 45–55MM for Disney+) and averages the highest watch time per subscriber, too. Netflix also has the highest share of direct sign-ups too (98% compared to the 69% market average). Because of its scale, Netflix is also believed to pay platforms such as Roku one-time bounties for sign-ups, rather than an ongoing portion of subscriber fees. This is particularly rough because Netflix has the highest ARPU in the SVOD market, not just the most subscribers. In addition, 100% of Netflix’s non-direct subscribers come from iTunes, versus the market average of 32%. iTunes is a platform, but means that if you’re watching on Roku or Fire TV, the platform fee doesn’t actually go to the one providing the service. In sum, Netflix is disproportionately popular and  disproportionately watched, but while it is disproportionately pricey, it’s also disproportionately sold direct or outside the connected TV device ecosystem, and pays disproportionately low fees. Not great! And not the sole instance of revenue leakage. Less than 4% of the audience for YouTube, the most watched streaming platform globally, even subscribes to YouTube Premium. In other words, the two most watched streaming services generate almost no revenue for the devices that are used to watch them.

To make matters worse, the fees these platforms charge have been falling for years. At the start, resellers took a standard 30% cut before shifting to “30% in year one of a subscription, 15% in year two,” which effectively became 15% flat. But since then, rates have fallen ever further. WarnerMedia struck a deal for particularly low rates on Amazon in exchange for renewing a large AWS deal. And since 2019, Apple has operated a premium partners plan that enabled select providers to pay 0–5% for select transactions.

This decline made sense. As mentioned, it used to be a challenge for a consumer to get streaming content on their TVs, and many households didn’t know how to purchase access either. Roku and Fire TV devices solved the first problem, facilitated the second, and simplified the overall process. But for years now, this has been relatively easy and deeply familiar to the average purchaser. What’s more, the average household likely has a few devices kicking around that can stream video to their TV—and since 2015, half of all new TV sales in the United States have been smart TVs that don’t need a standalone Roku or Apple TV device to stream video apps, which means that Samsung, Vizio, LG, et al. are now competitors. These many changes diminished the power of smart TV operators as a whole, forcing them to compete more on pricing. Disney+, for example, did not come to terms with Amazon’s Fire TV until five days before the service launched, and it took more than a year for Peacock to launch on Roku and Fire TV. In the mid-2010s it would have been impossible to operate a streaming service that wasn’t supported by those devices.

In January 2019, 75% of resellers were iOS (30%) and 46% Amazon Channels. By the end of 2022, these two services had lost fifteen percentage points of share and held an even smaller share of new subscribers.

Another driver of declining reseller margins is the decline of the “Channels” model that was pioneered by Amazon in 2015. Prior to Amazon Channels, the typical platform reseller model worked as follows. A user would sign up for a given SVOD service on a given platform (Roku, iTunes, Android, etc.), and the platform would manage the billing in perpetuity. But the user would still have their own log-in credentials for that service and could download that service’s app directly on other devices, so a user might subscribe to HBO on Roku but then also watch HBO on a Fire TV or iPad using their HBO credentials. Under the Amazon Channels model, all consumption would happen on Amazon irrespective of the platform used, so HBO would be watched within Prime Video. There were several benefits to the Channels model, such as the fact that the underlying SVOD owner did not need to manage a complex video streaming stack, ensure compatibility across all end-user devices, or cover customer service costs or bandwidth. In effect, it allowed a cable network to keep being a cable network! All they had to do was make, release, and market their content. Customers liked this model too, as it meant that all of their content, credentials, and payments were in one place. As a result, most SVOD operators found that subscribers who watched their services via Channel subscriptions, rather than standalone apps, were cheaper to acquire, more engaged, and less likely to churn.

However, many SVOD operators eventually considered this model a long-term competitive threat. In addition to the extra 10–15% in commissions (or 30–40% total), the Channels model meant that an SVOD was effectively absorbed by the Channels platform. In other words, a participating network would not have their own UI/X and not have direct access to the customer’s personal information or engagement data. In a Channels context, their content would also be seamlessly folded into all other content. An HBO show and a Showtime show and a Starz show and the platform’s own original content might all be side by side and without clear branding distinctions. As the “Streaming Wars” progressed, these drawbacks felt increasingly problematic. Meanwhile, the streamers were getting better at customer acquisition, video delivery, customer service, and other aspects of technical delivery.

As such, all new entrants after 2018 skipped the Channels model, including Apple TV+, Disney+, Peacock, Quibi, Discovery, and Roku Channels, while HBO left the Channels system in late 2021, and in doing so, lost nearly 5MM subscribers it could only hope to reacquire elsewhere. The services which continue to use Channels systems are typically weaker and/or niche SVODs, such as those of AMC Networks as well as Showtime and Starz. Six months after Discovery’s acquisition of WarnerMedia, as the newly formed entity contended with record leverage (4:1 debt/equity), $5B+ in write-offs, and a more than 50% decline in share price, the company announced it would return to Channels programs. Regardless, the return of HBO Max is unlikely to reverse the model’s decline. At the start of 2019, 47% of all SVODs billed by a reseller were via Channels programs, with up to 60% of new sales going through that model. By the end of 2022, these figures had fallen to 30%.

Connected TV device platforms do generate revenue from third-party streaming services beyond taking a share of subscription revenue. Specifically, they take a portion of the total ad inventory from ad-supported and AVOD-only streamers. However, this specific share is variable. Roku’s distribution agreements require 30%, but it’s widely believed that many are less and that many other top SVODs, such as Disney+, Netflix, and YouTube, hand nothing over (as of January 2023, Disney+ with ads is not available on Roku, almost certainly for this reason).

The challenge with falling rates and non-revenue-generating consumption is that Connected TV platforms adhere to the razor-and-blades business model. The physical boxes are typically sold for less than they cost to produce, and while an increasing share of Connected TV platform users do so purely through software that’s baked into their TV, Amazon, Roku, Apple, Google, and others compensate TV manufacturers for this placement (they pay even more to be the only option, and more still to replace the OEM’s operating system outright). This strategy is based on the fact that consumers are highly price-discriminating when picking a Connected TV device and largely indifferent to the software a smart TV uses—but the platform that is selected benefits from recurring high-margin revenue streams (subscription commissions and a share of ad inventory) that can last for years. Thus far, however, this business model has not proven to be lucrative. Since its IPO in Q1 2017, Roku has recorded a quarterly profit in only 6 of 22 quarters, most of which followed the 2020 wave of lockdowns. During the first three quarters of 2022, net income was −$280MM on revenues of $1.5B.

Although Connected TV devices are still figuring out their business model, it’s clear their model beats Google’s original Chromecast strategy. For a time, Chromecast seemed to be a major disruption in access innovation. At launch, it was a third of the price of Roku and a tenth of Apple TV. The device was also a fraction of the size - so small it could be plugged into and hidden behind a television, rather than set on top of a media console. Unlike its competitors, it didn’t require learning a new device, managing a new operating system and loading it with apps (that a user then needs to sign into), or a standalone remote. Instead, users would just send the device instructions or video data from their smartphone, which the Chromecast would then relay to the living room screen. This was a more elegant approach from a technical perspective, one that utterly bypassed extra steps and systems and accounts and apps. It didn’t provide Google with the ability to take 15–30% of transactions—at least to start—but it lost far less on hardware, collected data anyway, and had the ability to place ads before streams. However, even this proved too difficult for the average consumer. The extra steps required by alternative devices actually made things simpler for the end user. And so in the years after Chromecast’s debut, Google shifted to what’s now known as Google TV, which is effectively identical in approach to Apple TV, Roku, and Amazon Fire.

Commercial and technological “settlement” happens during all access waves, irrespective of the dominant business model, technology, or company at the time. For years, Spotify and Pandora, two different visions of audio’s digital future, topped the app store charts, while executives at each company (and the major labels) suggested that on-demand listening, as well as interactive radio, could and would exist together as successors to terrestrial broadcast radio, satellite radio, and download-to-own stores. Steve Jobs famously believed that neither model would replace the download-to-own model. But eventually, we conform. Apple launched its own Spotify competitor in 2015, while Pandora launched its own in 2017. The former service thrives and has long surpassed iTunes in listener hours, while Pandora’s service, arriving too late to see much adoption, was shut down. Pandora’s core offering, interactive radio, has seen its monthly active users decline nearly every quarter since Q4 2014, from 82MM to 49MM, even though streaming audio revenues have grown from barely $1B to more than $10B per year during the same timeframe. Meanwhile, Spotify, Apple, and Amazon, having added interactive radio (via features such as Discover, Playlists, curated mixes, and the like), have netted over 120MM users in the United States alone, with each user averaging four times the monthly listening time of the average Pandora user.

But to return to streaming, the market has broadly centered on SVOD and adopted many of the tactical models pioneered by Netflix. One such example is “SKU collapse.” Even seven years ago, the consensus future was “an app for every channel” and an SVOD for every genre.

There were good reasons for myriad networks in the pay-TV era, or most of it anyway. Linear distribution meant that a single channel could only air one program at one time. As such, there was only so broad a title could get before alienating some viewers who would want more comedic or more dramatic stylings, more or less graphic violence, a more broad or specific tone, etc. The only way for a company to cater to everyone was to launch multiple channels catering to many groups of them. A streamer, meanwhile, can offer all types of content to all people at all times—a far better offering for the consumer (and especially the family). And so that’s where we converged.

At one point, WarnerMedia had nearly 10 SVODs (HBO Now, Cinemax, FilmStruck, Warner Bros Archive, DC Universe, Crunchyroll, Drama Fever, Rooster Teeth) with plans for more (TBS, TNT, Turner Sports, CNN) before consolidating them into HBO Max. NBCUniversal used to have standalone TV Everywhere apps for NBC, USA, Bravo, CNBC, E!, Syfy, etc. Now all of NBCUniversal’s network-specific apps are just cosmetically reskinned versions of the same app, with access to the catalogues of all of their sister networks, each under a branded section. Similarly, NBCUniversal has only a single SVOD service today, Peacock, which contains all of its comedies, dramas, films, documentaries, and whatever else.

Disney+ could have been four different SVODs (Disney Pre-K, Disney Teen, Disney General Entertainment, and National Geographic). But collapsing them into a single SKU made for a more compelling service with better ARPU-to-churn (i.e., LTV) dynamics, even though the single-SKU model means some households pay only $10 a month for Disney+ where they might have spent $20 for all four. Hulu and ESPN+ remain outside Disney+ today in the United States, but abroad, these services are typically into a single service. For example, what Americans consider to be “Hulu” is just a tab inside Disney+ in Canada, the UK, and South Korea, under the name Star. In other markets, Disney+ is folded into Star+ alongside live sports content. Many believe that after Disney buys out Comcast’s remaining share in Hulu—or Comcast buys out Disney—the service will be folded into Disney+ or Peacock.

Another lesson from the Access Era was the criticality of a global footprint. Hollywood networks have always had foreign channels. In 2014, HBO could count nearly 100MM subscribers, a feat Netflix didn’t pass until mid-2017. Yet the nature and especially the economics of these networks varied greatly. Outside of the United States, HBO received as much as $13 per subscriber per month (Canada) and as little as $0.14 (parts of Southeast Asia), with the network operating as a premium cable network (as it does in the United States), basic cable network (akin to FX in the United States), or even broadcast network (such as NBC). Sometimes HBO simply licensed its library to a foreign network (e.g., Canal in France) or its library plus the right for this independent network to name their channel “HBO”. This variability reflected several realities. Many markets, such as Japan and India, had little to no cable penetration, blocking both the basic and premium cable model. This meant that if an American network want to operate their own network in these markets, they had to secure public broadcast rights—should they even be available, let alone to foreign entities—and build a local ad sales force to support a purely ad revenue–based network. Foreign operations were also laborious, requiring extensive investments in a local office, sales force, marketing, distribution agreements with local cable providers, bank accounts and accounting, and on and on. The flexibility of Hollywood’s international models allowed each player to avoid the pitfalls of individual market dynamics, scale, price points and currency fluctuations, while matching their own priorities, capabilities, and investment abilities.

In SVOD, global distribution has become a key tactic. Following its 2016 expansion, Netflix was available in nearly every country, territory, and market globally, including Cuba, albeit with the exception of South Korea. That same year, Prime Video expanded nearly as widely. Within 12 months of its 2019 launch, Disney+ was available in more than 75 countries and now reaches nearly 150. Apple TV+ started with over 100, and Paramount+ launched in several dozen before setting its sights on passing 150. During this same period, HBO began buying out its foreign partners, taking control of HBO Asia, as well as HBO Latin America, while launching OTT-only service in markets such as Norway.

Through over-the-top delivery, any network, based anywhere, can, quickly and without considerable cost or effort, launch their services worldwide. They need not secure public airwaves, find local pay-TV providers to syndicate their content or channels, seek regulatory approval, or build a local ad sales force. Of course, local content would be key to scaling a local subscriber base, but taking on modest incremental programming expenses and operating costs was well worth the ability to directly monetize the end viewer rather than be intermediated by a local player. Furthermore, every international subscriber helped to amortize “domestic” content investments and global tech investments, and every foreign-language series had the opportunity to become a hit in the English-speaking world (Squid Game, Dark, Money Heist, etc.). Operating internationally also helped defend nationally. If Paramount+ wasn’t available in Latin America or the Nordics, it would struggle to afford many of its signature, US-centric series—unless it licensed the foreign rights to services such as Netflix or Amazon, with whom it competed in the United States.

As many “legacy players” have long known, ads are important to the future of every streaming service. Users, of course, would rather have ad-free video. However, the very fact that the dominant streaming video services (Netflix, Prime Video, Disney+) used this preference to accelerate adoption and cannibalize viewing time from the typically ad-crammed pay TV offering intensified the business case for advertising. Put another way, the proliferation of ad-free OTT services meant that premium OTT ad impressions were incredibly scarce, which meant they were incredibly valuable to advertisers. Still, ads are inevitable because as much as consumers might dislike them, they won’t pay enough to avoid ads relative to the monetary value of watching them. This is demonstrated by the fact that ad-supported tiers typically generate more revenue per user than ad-free tiers, even though they charge the user much less. The best example of this was Hulu’s 201[x] decision to cut its ad-free price from $8.99 to $5.99. This “cost” Hulu more than half a billion in subscription revenue per year, but it widened the gap between ad-supported and ad-free (then $11.99). This, in turn, drove many happy ad-free customers to downgrade their service in order to save what was $3 per month but had since doubled to $6. At the same time, Hulu was generating $10-11 in advertising revenue per user per month, meaning that the extra $3 or $6 lost still netted out positively. And that was then! CPMs have increased since 2017, but more importantly, ad revenue is highly scalable with engagement time. The more you watch any specific streaming service, the less likely you are to churn. Thus, does marginal consumption drive business value. At the same time, engagement, arguably the most important user metric for an OTT video service, quickly encounters diminishing (often effectively zero) marginal returns. Most users will resubscribe if they’re watching 30 hours every month; growing to 60 might increase renewal rates by only a few bps. And as services are “all you can eat,” these customers generate no revenue from this consumption while simultaneously chewing through content that would otherwise keep them a subscriber for months to come.

The one shortfall with the advertising model is that it’s harder to scale internationally. Subscription revenue is simple. Foreign exchange rates do swing, but they can be hedged as the largest markets tend to be the least volatile. But most video advertisements are locally made and tailor specifically to that market. And while the largest advertisers globally are in automotive, fast food, and wireless, they don’t advertise the same price, same imagery, or even same products in every country.