Game Companies Are Not Tech Companies Part III: The Content Problem

So maybe game companies aren’t tech companies. As much as game companies seem to borrow from tech firms, tech firms run a bigger deficit in borrowing from games. If Netflix, a subscription service with over 10,000 movies and TV shows, has its biggest competitor in a single game, Fortnite, then perhaps there’s more for tech to learn from games. And how games deal with The Content Problem is its defining characteristic. Of all forms of entertainment, games present the most compelling answer to the problem.

The Content Problem

The fundamental axiom of economics is unlimited needs and wants and only limited means to fulfill them. The parallel for entertainment might consider that core content demand nearly always outstrips supply. For example, large swaths of Game of Thrones and Harry Potter fans are underserved by a couple of books, movies, and TV seasons. Executives try to fill the void with licensing: Harry Potter backpacks, Game of Thrones beer, etc. But filling the core content demands are impossible: it takes far more than 1 hour to produce 1 hour of Game of Thrones, while the same is not valid for games.

Consider the following:

The content consumed in a game like Overwatch or Clash Royale is the pursuit of strategy equilibrium and/or mastery of mechanics. A new unit in Clash Royale changes how players organize their decks, even if they don’t use the unit directly (they must counter it). The new units provide hundreds of new hours of content to consume relative to the near 1 man-week of labor to produce the new unit. Therefore, the marginal content output of a given member of the 16 people (!) Clash Royale team is astronomical.

The genius of PvP (Player v Player) environments is how they necessitate the emergence of a meta-game. PvP environments resemble game theory models where it has been shown strategies evolve in an evolutionary process. In mathematics, Player vs. Environment (PvE) resembles optimization where strategies are static – one and done. Each balance change reshuffles Equilibrium in PvP environments; the search for dominant strategies in an ever-shifting equilibrium is the game itself.

The marginal product of labor for a given game developer completely outclasses a given producer on a movie or TV show by virtue of the medium, not the individual. Unlike games where assets are infinitely replicable, films and TV face fixed constraints: Emilia Clarke or David Benioff can only be in a single place at a given time. They must also eat, sleep, socialize (sigh). Meanwhile, Captain Price faces no such constraints. There’s no more Game of Thrones to consume after the last episode cuts to black while there’s always another hour of Fortnite to play. How can Netflix and others adapt to the reality of these mediums?

The most straightforward strategy is a content arms race. Netflix continues to spend over $17B a year on original content while scooping up oodles of back catalog content. Of course, viewers must be interested in this content to be “effective,” and the recommendation engine plays a vital role in this. But the last episode of Stranger Things shows that the recommendation engine cannot fill the void while operating on the same indifference curve. The “more bodies” strategy to solving the content problem is expensive to execute and, as we’ll see in part 4, struggles to achieve Marginal Cost = Marginal Benefit. Reality TV is a response: it needs fewer writers, editors, and CG to produce a given hour of content. Shows like The Amazing Race, Big Brother, and Survivor can do 20+ seasons of 22+ episodes, while Game of Thrones struggles with seven seasons of 10 episodes despite having so many more crew members. Netflix’s speed of investment here is breathtaking. But the addressable audience is more limited in scope than traditional dramas. Netflix needs a bolder evolution to combat games: TV-as-a-service. The forgotten genre of soap opera TV provides a near-perfect blueprint. For those unfamiliar, soap operas are near year-round weekly serialized television shows. The unrelenting pace has resulted in popular series like General Hospital having 14,000+ episodes over 57 years. Netflix needs to invest in moving performances to a similar format: year-round production with weekly releases heavily. There’s always another piece of content to consume right around the corner while the back-catalog for a given show continually expands for newcomers. In many ways, this mirrors match-3 level production. The number one reason why players churn from match-3 is a lack of new levels, and a glance at community pages confirms this. King mitigates content churn using an increasing difficulty curve such that it takes players longer to reach content exhaustion as they progress through levels. Another strategy is also possible, however: branching narratives. Reality TV faces no such option. Increasingly, Hollywood is shooting movies back-to-back. It’s cheaper to continue production rather than stop and go. Why not do something similar to produce more content? In this case, shoot multiple perspectives in a given series simultaneously. Lord of the Rings production operated similarly with two production crews but with a singular end product. Game of Thrones also operated with two film crews, but the end product was a single episode. Why not dedicate an hour to each perspective rather than splice the two in a single episode? A multi-perspective production multiplies a 10 episode season to 30 while holding down cost. Netflix can’t solve The Content Problem, but it can mitigate it. Interestingly, Youtube has solved most of this problem via a two-sided marketplace. The smattering of volume helps the supply-side problem even if a particular creator has a finite number of videos (remember, you can still play a given game for an unlimited amount of time without “running out” of content). Youtube has encouraged users to subscribe to many different creators, accounting for regular release cadence. Diminishing returns for linear content are incredibly steep; few users will watch a film or movie more than once. Increasing and prolonging the LTV of a viewer is most elastic with more content: a costly proposition. To compete with games, TV, and movies need far more supplies. If technology and business models can change innovative products rather than be a vehicle for them, now has never been a better time to explore changes in storytelling. Part IV Game Companies Are Not Tech Companies Part II: Platform Power (Or Lack Thereof) Part I Ran Mo retraces the incredible fast follow of the auto-chess genre. Before you could blink an eye, what was a mod in Dota 2 (which itself was a mod) had spun off into three incarnations. The first was from the mod maker, Drodo Studios’ plainly labeled Auto Chess, the second was Valve’s Underlords, and the third was Riot’s Team Fight Tactics. It’s not clear any of them have done “well.” Indeed, none of them are making any real money as there’s little to purchase amongst any of them. Valve’s Underlords PSU Declines But if there is a leader, it’s Riot’s Team Fight Tactics (TFT) by content cadence (from a player count perspective, it’s unclear if TFT is doing something meaningful). How do we explain the last entrant now leading the pack? Ran seems to argue for the power of the platform: “Underneath the hood of League of Legends is a deep set of systems and tools: messaging and voice chat systems, social systems, matchmaking, internal development tools, event systems, cloud-based server infrastructure, post-game data analytics, and so on. These tools enhanced play experiences and supported the continuous release of contents, events, and updates.” Ran goes further, explaining that “We’re likely entering a new wave of consolidation today, with new moats centering on live-service ecosystems […].” I could not disagree more. Not simply on Riot’s platform power, but the wider value of game platforms altogether. Game platforms have, and will always be, valueless: there is too much game specificity and velocity for meaningful platform features to scale. “[…] messaging and voice chat systems, social systems, matchmaking, internal development I could not disagree more. Not simply on Riot’s platform power, but the broader value of game platforms altogether. Game platforms have, and will always be, valueless: there is too much game specificity and velocity for meaningful platform features to scale. “[…] messaging and voice chat systems, social systems, matchmaking, internal development tools”? While standard and required for live service, they are not particularly “deep.” If these features form a moat, then it’s the equivalent of a kiddie pool defending Versailles. The ever-growing number of PC game launchers, all of whom do these exact things, suggests as much. The primary value of game “platforms” is discovery and perhaps security. The deeper platform features of Steam (Marketplace and Workshop) have failed to see widespread adoption. Other platforms haven’t even tried building anything beyond a friends list and chat. None of them can justify a 30% cut, and for many games, “multi-tenancy” or listing on multiple platforms is the clear revenue-maximizing strategy. Of the top 20 Steam games, only 35% non-Valve games use either Marketplace or Workshop. Furthermore, 59% are multi-tenant or are on other PC launchers. Expect this to increase as Steam degrades as the sole game discovery launcher. Activision realized this long ago when they shifted Call of Duty to Blizzard.net and away from Steam. Blizzard titles anchor the launcher with eyeballs but do little feature work. It’s 2020, and Blizzard still region locks many titles, meaning switching from NA to EU servers entails losing all progression for some titles. Other firms take game platforms to eat at certain parts of the game “supply” chain. In addition to a given piece of technology working for games, these firms hope the tech will scale outside of games as well. This is an executive fetish that rarely, if ever, pans out. Machine Zone (MZ) is the latest failed entrant into “but we’re really a tech company”. Both Cognant, its internal ad-buying platform, and Satori, its cloud platform, went up in flames. MZ ended up selling to AppLovin at 10% of its peak value. The jury’s out if Improbable, the latest “but we’re really a tech company” can dig out of$85M losses.  A more considerable hope is Unity, but even then, only 8% of the 716 customers to spend $100,000 or more have been non-gaming firms—games as a technology an executive fetish that rarely, if ever, pans out. Epic’s PR team was out in full force trouting The Mandalorian using Unreal, but indeed this is a mingy share of Epic’s revenue. Games have similar challenges to tech firms, but tech firms rarely have identical challenges to games (all squares are rectangles, but not all rectangles are squares). A common theme of this blog reiterated in this series is that games are a distinct and unique medium. The faster we embrace this, the faster the industry can evolve. While game companies may never be tech companies, there may be a small place for game technology companies. Unreal and Unity are real and are here to stay – they solve complex and prevalent problems game makers face. If the games industry grows, the total addressable market for game tech solutions grows as well. Is there room for more game technology companies beyond Unreal and Unity? The previously mentioned Improbable seems to think so, but little additional evidence exists. There’s more to be said for a given game as a platform, but the evidence is scant even then. Publishers haven’t been able to retain the value of extensions within a given title (Auto Chess splitting from DOTA 2) or outside of it (MOBAs splitting from Warcraft III and Heroes of the Storm). The simple ability for players to build content within games hasn’t precisely accelerated either. If anything, modding has become more challenging, not easier since its ascent in the early 90s. Can we think of a single major 2020 release to embrace mods? Much was made of Halo 3’s Forge, but we’ve seen few copycats since its release in 2007. If we consider a corollary to Bill Gates’ definition of a platform, it might go something like this: “A game is a game platform when the playtime of platform content eclipses the main game.” In this view, there hasn’t been a modern game as a platform. A softer version might consider a game as a platform when “the playtime of additional modes eclipses the main game.” Despite a much lower bar, it’s not clear that Fortnite, with its concerts and social spaces, has reached it either. We’re left with Roblox, a trustworthy game platform. Roblox provides valuable tools that make it easy for developers to create compelling games and provide for discovery. But the jury is out to consider this the future of games, especially with its elder sibling, Manticore, is going nowhere fast. RPGs vary from CCGs, which vary from FPSes. FPSes might need to solve 64 player servers, while CCGs may need transfer markets and RPGs need deep customization systems. This specificity of these challenges shrinks the total addressable market for the tech solutions game firms have devised. Central platform features haven’t adapted to a particular game design (and the speed at which it changes). As a result, we’ve rarely seen game companies make a successful transition to a tech firm. Part III Game Companies Are Not Tech Companies Part I: Every Game Has Networking Effects, They Just Don’t Amount To Much Ran Mo and Joseph Kim (@jokim1) argue for looking at games from a Silicon Valley perspective. The usual three make an appearance: moats, networking effects, and platforms. Moreover, while I thought it had died out after the launch of Halo 3, there continues to exist an inferiority complex amongst game makers. Games never seem to get the mainstream or broader tech circle legitimacy many think they deserve. Despite operating in the Valley and major tech hubs, game companies do not reach the crazy evaluations of FAANG (ATVI has a market cap of$63B, while Facebook sits at 215B). Furthermore, public intellectuals like Tyler Cowen or Ben Thompson rarely discuss games as a share of their commentary (not the case with Matthew Ball – but maybe he drinks too much kool-aid). Even internally, game firms seem to be chasing subscriptions off the heels of Netflix and Spotify. The fundamental problem comes from not respecting or understanding games as a distinct How to Build the Amazon of Game Companies mainly describes why game companies will not be Amazon. The fundamental problem comes from not respecting or understanding games as a different and unique medium separate from linear content. In many ways, Kim’s piece wants to make this point but does not go for the kill by the end of the article.. I’m writing a some posts to address and expound on this. In two parts (networking I’m writing a two-sided series to address and expound on this. In two parts (networking effects & platform power), I’ll examine why game companies fall short of traditional tech companies. Then, with another two parts, I’ll address what tech companies have to learn from games (the content problem & MC = MB). Networking Effects Networking effects describe the positive externalities from the n+1 user to a service. Every time someone joins Facebook, the benefit increases in value as others can interact with that person. Gyms, for instance, work in the opposite direction: every member who joins a gym occupies a fixed amount of capital and decreases the value to each other member. Fawning over networking effects comes from the path dependency inherent in the model: more users leads to more users. What does VC not want self-perpetuating growth? However, as Margolis and Liebowitz argued during the Microsoft case: Although these simple numerical and algebraic examples appear both logically sound and structurally uncontroversial, these examples entail severe restrictions. The logic underlying path dependence is seductive but incomplete. […] Given that the theoretical claim that can be made for path dependency should be understood as only a demonstration of possibility, the case for path dependence becomes an empirical one. It is not that networking effects are not real, but they are not as powerful as they were first made out to be. After all, networking effects could not save Friendster or Myspace, and as we will see, they mean little for particular games. At a certain scale, diminishing returns decay positive network effects to zero. The 2nd user who joined Facebook was far more beneficial to the 1st user than the 100th million users who joined. While not directly comparable, Google’s Chief Economist, Hal Varian, makes this point regarding the predictive accuracy of models with additional data. And we can create a similar arbitrary model for a particular game: diminishing network effect value for the marginal user that inevitably results in an asymptotic total network effect value. Not all games face the same curve. A game like Hearthstone, with 1v1 play, has far less to gain from an additional user than, say, League of Legends, which has 5v5 and many ranked segments. More users reduce matchmaking times, potential latency, and thickening skill distribution (higher P, you will be matched against similar skills). The fewer segments (modes, ranks, etc.), the less powerful networking effects are, and the quicker the marginal value curve depresses. Cross-play doubled down on this by removing platform segmentation. Nevertheless, even for League, the networking effect power is infinitesimally small at scale; the significant gains are eaten with a relatively low user count. Games do not have the “sticky” elements of networking effects. Synchronous consumption is another example. In any real-time game, a network effect is delivered when you play with friends. At the same time, something like Instagram Stories is consumed whenever the user pleases. It is not clear that having a friend play the same game I do is beneficial unless we play simultaneously. Schelling’s Nobel Prize-winning work on tipping is a more apt model for describing many games today. Consider a group of players all playing the same game. These players are partially driven to play this particular game to be “in the know” or a part of the pop-culture conversation. Each of these players has a given threshold for defecting to a new game based on the share of public discussion consumed by the game. Players defect if the game declines as a share of the conversation; they leave as the game goes from 100% of the public conversion to 99%. More will leave at 99% and more still at 98%, continuing a downward spiral into a new equilibrium. Of course, the inverse is true as well. Some might join a game if its share of public conversation goes from 0% to 1%, and even more would participate if it went from, say, 20% to 30%. A new game release can set off this “tipping” chain reaction in players. Look no further than the migration of Fortnite players into Warzone. The power of tipping is as strong as “public conversation” is as player motivation. Unfortunately for developers, this means instability in the long-run capitalization of viral game hits. I have avoided addressing two-sided marketplaces as they will more neatly fit into the next part: platform power (or lack thereof). Part II Can We Get Players to Tell Us Their LTV? Eric Suffert acutely describes the dangers of extending payback windows. At every t+1 the accuracy of LTV declines while the variance in cohort profitability increases. LTV, however, is not an exogenous variable and clever design can incentivize players into revealing their long-run time horizons within a game. Consider the design of a many subscriptions: you can pay a lower annual fee or a higher month-to-month fee. If you’re uncertain about the subscription, then the month-to-month is more economical while if you have more certainty then the annual fee makes more sense. The choice is a huge predictor of retention: annual users are far more likely to retain then month-to-month users. The mere inclusion of this annual/month-to-month choice gives users the opportunity self-segment into more predictable cohorts. Why can’t we use the same mechanics in game design to create more predictable LTVs? Consider two possible goods for purchase via gems in Clash of Clans: a builder or gold. The builder increases the long-run growth rate of gold while the gold itself is a temporary boost in short-run capital stock. In layman’s terms: spending 100 gems on a builder might net you 200 gold today and 1,000 gold by D30 while spending 100 gems directly on gold may only yield 700 Gold today and 0 gold by D30. The builder is an annuity that pays dividends every period, the longer a player’s time horizon the more valuable the annuity. Players who expect to have a long time horizon in a given title have an enormous incentive to purchase “investment” goods or goods that pay dividends overtime (battle passes similar to some degree). Not doing so results in a increasing opportunity cost penalty every period due to lost compounding growth. F2P has experimented with direct daily annuities of hard currency. They offer players a discount over the standard IAP packs but must pay upfront to receive a daily allowance. Instead of a 30-day pass, why not ramp to a quarterly or bi-annual pass? Doing so would make LTVs more predictable early in given player’s lifecycle. Kantian Ethics for Game Ads and Beyond Is Probably a Good Idea The ASA has banned misleading ads from Playrix’s Gardenscapes. Running the same creative in user acquisition hits diminishing returns fairly quickly as the creative “clears the market” for users attracted to that creative. To broaden appeal, why not simply advertise the game as existing in a entirely different genre? This opens up a whole new segment and drops CPIs significantly. Of course, advertising gameplay that doesn’t exist surely means these users will exhibit extremely poor KPIs. However, there’s a broader implication to these ads and one that harms the entire game industry. The harm is described in Akerlof’s Nobel Prize winning paper on car lemons: Akerlof’s original example of the purchase of a used car noted that the potential buyer of a used car cannot easily ascertain the true value of the vehicle. Therefore, they may be willing to pay no more than an average price, which they perceive as somewhere between a bargain price and a premium price. Adopting such a stance may at first appear to offer the buyer some degree of financial protection from the risk of buying a lemon. Akerlof pointed out, however, that this stance actually favors the seller, since receiving an average price for a lemon would still be more than the seller could get if the buyer had the knowledge that the car was a lemon. Ironically, the lemons problem creates a disadvantage for the seller of a premium vehicle, since the potential buyer’s asymmetric information, and the resulting fear of getting stuck with a lemon, means that they are not willing to offer a premium price for a vehicle of superior value. If users start to have an expectation that a given game ad does not truthly describe the game in question then they’ll be less likely to click. This means higher UA costs even if your firm does not engage in these type of ads. A similar problem is starting to creep up in PvP games. Developers have started to be confronted with the uncomfortable reality that PvP is a zero-sum game: for someone to win, someone else has to lose. And of course, when players cannot progress they churn. Supercell has heavily introduced bots in Clash Royale as a response. Of course, players cannot identify if they are indeed playing a bot or a real player. This steals one of the compelling aspects, if not the compelling aspect, of PvP away from players: outsmarting another individual. But like the lemon problem above, if this trend continues then players will start to question if they’ve dominated a real opponent even if the particular game doesn’t use bots. Games that use unmarked bots start an industry expectation that diminishes the experience for all. There’s a good moral rule here to help us and it’s from an 18th century philosopher called Immanuel Kant. Kant advocates for something called the Categorical Imperative. This claims that if we were to universalize a given action and it would result in a “contradiction” then that action is immoral. Consider lying: if everyone were to lie then the world would not function, therefore lying is immoral. Or consider being lazy: if everyone were lazy then nothing would get done. It’s a no-holds-barred approach to consider moral action, but thankfully we’ll use it a much more narrow scope. If unmarked bots were to be universalized, PvP games would be irrevocably harmed. If misleading ads were to be universalized, then players would stop clicking on them all together. Both of these situations violate the Categorical Imperative and align with outcomes that benefit the entire industry. The Categorical Imperative makes for a simple and rule based approach to consider the “greyer” parts of developer action. The Collective-Action Problem of F2P Clans Remains Unsolved There’s a compelling aspect to achieving group oriented goals: being apart of something larger than yourself. Lots of F2P developers harp on the importance of social features. Yet the social experience in many games is abysmal. Lots of teammates or clanmates don’t seem interested in participating instead preferring to “free-ride”, putting forward little effort but getting the fruits of the team reward. Mancur Olson’s foundational work, The Logic of Collective Action, describes how this problem manifests in the public sphere (sometimes literally in the case of electric scooters). Game designers have a much easier time aligning individual and clan incentives than public officials yet they sometimes miss easy wins. How can we make the clan experience better then it might otherwise be? In Clash Royale, clans advance a boat against rival clans. Advancing the boat depends on individual clanmates playing games everyday (and winning). The more clanmates play consistently, the more the boat advances and the better the rewards the clan will receive. But for many clanmates playing everyday requires a great of effort, why not let others earn the rewards for you? The problem is severe in Battlefield where “PTFO” or “Play the Fucking Objective” is standard nomenclature. Players often won’t engage in activities that benefit the team (capturing flags), instead preferring to pursue their own objectives (generally: shoot players as fast as possible). A given player faces two potential payoff schedules when considering to allocate effort to the clan. There’s the expected payoff with no effort (the probability that the clan/team will win if the given player did nothing) as well the probability that the clan will win if the player puts forth effort. We can model this as such: $\mathit{expected\ payoff\ from\ effort_i} = {P(\mathit{winning} | \mathit{effort_i}) \ *R}$ where $P(\mathit{winning} | \mathit{effort_i})$ is the probability of winning the clan event given give the effort of a given player or rather the additive probability of this given player participating. While R is the reward from winning. Of course if $$P(\mathit{winning}| \mathit{effort_i}) = P(\mathit{winning})$$ or the given player cannot sufficiently contribute to the probability of the clan winning then there’s zero incentive for them to put forth effort. Why bother? This problem exacerbates as team size grows: the efficacy of a given player varies inversely with the number of teammates. This makes intuitive sense: in Battlefield, a player in 2 versus 2 match has a greater impact on the outcome then a player in a 32 versus 32 player match. The incentive to free-ride rises as the number of teammates or clanmates rises. Weakness hides in numbers. We’ve also ignored the game-theory dynamics of this problem for simplicity, but it’s worth mentioning. If I know my other teammates are not going to put forth effort, why should I? This leads to Nash equilibriums where clans have almost no activity. How can we overcome the free-rider problem and ensure that all teammates put forth effort? The highest cost-benefit feature is simply better monitoring tools. In many clan or team based games, clan leaders face asymmetric information: they simply can’t identify the players that do not put forth effort. A simple measure of activity (last login) or games played in the last week goes a long way to kicking out free-riders. We might also consider a joint-production function. In Battlefield or Clash Royale each player would receive a score based on their effort or contribution to team advancement, if the team wins they receive a multiplier on this score. Such a system would have two benefits: it would more closely align individual effort with individual outcome (reap what you sow), and it would increase the benefit for high performing clan members to engage in monitoring. For example, a high performing member might have20 in contributions with a 2x multiplier or $40 for winning compared to a low performing member with$5 in contributions and therefore $10 for winning. In real terms, the high performing member has an even greater incentive to encourage low performers to put forth effort. There’s a lot to be said for social shaming as well. While it hasn’t been effective for zero effort participants, there’s evidence it might help players on the margin. A push notification demonstrating that your clans needs you or perhaps better yet, a system where your clanmates can send you push notifications is a compelling way to push players into action. Perhaps the greatest miss I see is not in clan monitoring (kicking out free-riders), but in self-selection to begin with. Clans are generally pareto efficient for players meaning that there’s zero cost and only benefit to joining one. Players then generally look for near max-size clans as they maximize the clan’s probability of winning a reward and thus the players. Reducing search costs by recommending (or restricting) clans based on device language, location, and some measure of progression maturity makes all players better off. It’s hard for social monetization opportunities to take-off if team based activities suck. We still have a long way to go to fix top of the funnel problems. Afterall, teamwork makes the dreamwork. The Economics of Game Pass Demand a Playstation Launch The supply-side economics of subscriptions are fairly straightforward: get massive scale and distribute costs. Netflix has over 160M subscribers meaning a$100M production only costs each user $0.63. There’s zero marginal cost (unlike Spotify), so the more subscribers Netflix has the less a piece of content costs on a per user basis. At$11 per user per month, Netflix can spend over $1.7B on content each month and break even. In fact, as recently as 2019, Netflix does spend close to break even. In the long-run, Netflix has a incredible loop: content brings users, these users lower per user content costs, this accelerates more content spend, which brings more users. Rinse, wash, repeat until diminishing returns take hold. But remember this wasn’t always the case. Content costs are fixed: the crew of Mindhunter doesn’t care how many subscribers Netflix has. They face costs unrelated to if 100M or 10M people watch the show. As early as 2010, Netflix only had 20M subscribers, suggesting a per user cost of$5 for a $100M production. There’s a gaping hole between$5 per user per $100M production and$0.96 per user per $100M production under their current 160M subscribers. This underscores the massive upfront investment needed for streaming to work on a supply-side basis. It’s this exact reason Microsoft needs Playstation users for Game Pass to at least have a chance at buying down per user costs. It also explains their aggressive sub & console bundling as well as rock bottom introductory pricing. Microsoft has disclosed Xbox Live has 90M MAU while Game Pass has 10M MAU. At$10 per user per month, Microsoft has can spend as much as $100M in content per month and break even. Development costs vary widely, but let’s start with assuming a conservative$70M per game. From a break even perspective, that’s around 17 games a year and consistent with Xbox Game Studios current production. That may sound like a lot, but the Xbox One has had over 2,500 games released in its lifecycle or about 350 games a year. Supporting this sort of volume is going to require a lot more users. A lot more.

Let’s have a little more fun with the cost side to get a better handle. Consider a simulation assuming anywhere from $30-$70M per game and anywhere from 100-350 games per year. We’ll randomly sample numbers in both ranges, multiplying them together to arrive at a per year content cost (ex: our simulation might pull $40M a game and 125 games for a yearly content cost of$5B. The density chart below represents 100,000 such pulls).

Simulating Content Costs of Game Pass

Based on this, Microsoft would most likely spend ~$7B per year to maintain similar volume. At$120 per user per year, this means 53M subscribers or about 5x the current amount to break even. At a 60% profit margin the number of subscribers jumps to 93M or more then the current Xbox Live MAU. Microsoft needs more users for this to work and they know it. Of course, perhaps this volume is insane and Microsoft wants much fewer but higher quality hits. In this case, however, the entire cost-savings model of subscriptions to the consumer declines. Our gaming subscription paradox re-emerges: subscriptions make increasing sense to the consumer wherein they would instead pay for many distinct pieces of content. Like music. Or T.V.

In many ways this explains the expansion of Game Pass to PC, Project xCloud, and the new Xbox console subscription bundle. Subscribers are capped by those with a console and Microsoft wants to start to unshackle that cap. Netflix recognized this long ago with Roku, Smart TVs, and apps on every platform or hardware they could find. The ability to sell Game Pass on iOS or Android explodes the addressable market, but is dependent on Azure wrangling the explosive unit costs of streaming. Hello Sony and the Playstation user base.

Microsoft clearly understands the supply-side economics of making streaming work, but over and over again misses the fundamental difference in how players consume a game as opposed to a T.V. show. I’m not quite ready to short Microsoft, but critical questions remain in executing on this strategy.

The Intellectual Poverty of Game Streaming & Subscriptions

Amazon has announced Luna, yet another stab at game streaming and subscriptions. It seems like the failures of Apple Arcade, Stadia, Gamefly, and our oft-forgotten OnLive have not been effective deterrents. Not to be outdone, Rovio’s Hatch continues to drain money every quarter. Perhaps this is what Bezos warned shareholders about when discussing new “multibillion-dollar failures.”

And despite overwhelming failures, media pundits like Matthew Ball prop up skin-deep arguments in favor of game streaming and subscriptions. Instead of discussing why game streaming and subs might work, let’s talk about why they haven’t worked.

It’s important to understand the brother/sister relationship between streaming and subs. Subscriptions unlock zero-marginal cost content consumption. Once you’ve paid Netflix or Spotify $10, there’s$0 additional explicit cost to consume another movie or song. However, there are transactions cost. In the “before times,” customers had to mail DVDs back to Netflix to receive the next DVD in their queue. This effectively limited how much content customers could consume in a given month. If mailing took 3 days in transit, on a 1 DVD at-a-time plan, a customer could only consume 10 DVDs in a 30 day month. This assumes the customer turned around DVDs instantly. Furthermore, the “queue” forced customers to plan consumption habits rather than at the point of consumption. If you wanted Love Actually on Friday, but by Monday, you were in more of a Pretty Woman mood, then you’re shit out of luck. Let’s not forget that new releases were in strong demand, meaning it could be weeks before Transformers 3 lands in your mailbox. Steaming solved all this.

Non-steaming subs like Game Pass and EA Play exist in a weird middle, solving some but not all of these issues. Games are distributed digitally, but not instantly. A game like Call of Duty: Modern Warfare can take 3-4 hours to download on a 135 Mbps connection. SSD’s aren’t cheering at the prospect of 200GB games either. But once streaming takes off for games, these problems are solved. No storage is needed!

The incredible rise of free-to-play and GaaS (Games as a Service) render subs and streaming largely valueless to the player. Players do not consume games like TV or music, which should have been made obvious in the last decade. Players are playing fewer titles in a given year but are playing fewer titles for longer periods of time. Games solve the content problem in a way that other mediums simply can’t.

The content consumed in a game like Overwatch or Clash Royale is the pursuit of strategy equilibrium and/or mastery of mechanics. A new unit in Clash Royale, for instance, can change how players organize their decks, even if they don’t use the unit directly (they must counter it). This can provide hundreds of new hours of content to consume relative to the near 1 man-week of labor to produce a new unit. Therefore, the content output of a given member of the 16 person (!) Clash Royale team is astronomical. Compare this to the thousands of crew members and weeks necessary to produce even a single one-hour episode of Game of Thrones. Supply can’t keep pace with demand in the world of TV and movies. Netflix makes sense in this view because after binging 9 seasons of The Office, customers can immediately rip into 7 seasons of Star Trek: The Next Generation. It’s another reason why back catalogs are so much more important to Netflix than they are to something like Game Pass. If players are only investing in 3-4 new games a year, then the transactions cost reduction streaming is extremely minimal the benefit it provides in high unit consumption TV and music.

It’s a similar story for the failure of gaming subs. If players consume only 3-4 titles a year, subscriptions don’t make economic sense. Not to mention these 3-4 new titles are increasingly becoming free. In the West alone, League of Legends, Fortnite, Apex, Warzone, and Rocket League dominate playtime. And let’s not forget the entirely F2P ecosystem of mobile. The march to F2P in the West will continue as long as MTX revenues grow and box revenues shrink. There isn’t a whole lot to save by signing up for a $100 a year sub and streaming service when Fortnite doesn’t cost a dime. Game streaming and subs don’t solve billion-dollar problems for the player. In the absence of doing so, subs and steaming will continue to flounder. The LTV-UA Rebate from Platform Fees Apple is increasingly under fire what’s claimed to be unfair practices in the App Store. The criticism takes three forms: (a) Apple’s 30% fee is much too high relative to cost (b) the rules are arbitrary and stifle competition and (c) the App Store as the exclusive avenue to install apps on iOS is unjust. That’s a lot to chew, so let’s focus on (a) for this post. Eric Seufert’s criminally underrated podcast talks about this very topic but phrases the question as “Does Apple earn it’s 30%?” Various App Store benefits like payment handling and preventing fraud are discussed, but I wanted to scream by far and away the biggest way Apple “earns” it’s 30%: acquiring a mega fuck-ton users to iOS. Nearly the entire value of a platform to a developer is how many people it can reach. No one is rushing to get on Epic Game Store (EGS) or develop for Stadia because there are so few users. 12% or even a 0% fee is irrelevant: [88% * 0 users] =$0 versus [70% * more then zero users] = more then zero. This is an extreme example, but users are by far and away the most important ingredient of any platform. After all, developers can list in multiple stores with maintaining the listing or code for the particular platform as the only cost to do so. The fact that so few are willing to take on these small costs tells us a great deal about the user bases of Stadia and EGS.

But platforms fees also increase the LTV of the devices who run on the platform, if and only if, the firm internalizes those platform fees. For instance, the added platform LTV of an Android phone in China is less than in the United States because Chinese users install non-Google owned stores. In this case, platform fee revenue doesn’t accrue to Google.

If LTV of the iPhone X is say $1000, Apple should spend up to$1000 in UA to acquire a user. However, with platform fees the LTV grows. Sensor Tower estimates that iPhone users spend over $79 per year on apps. On a lifetime basis, that’s probably north of$250 (3+ years of ownership). That ups the iPhone’s LTV to \$1250 and unlocks more UA budget for Apple to acquire more users. These marginal users benefit all developers. This is how Apple (or almost any platform holder) gives developers a rebate on the 30%. I’m not sure what the true fee is but it must be lower than 30%.

It’s Not Data-Driven or Informed You Want, It’s Science

“We want to be more data driven” or “We want to create a stronger data culture” are common organizational refrains. Supposedly, having more data or data playing a larger role in the decision making process is profitable. It’s weird because I haven’t seen any research to suggest this is the case. In firms like Facebook, it’s obvious as more data improves ad personalization and thus revenue. But this is data as a engineering project rather then a tool in the decision making process. Firms want to make better decisions with data. This is a misidentification of the value chain. Data isn’t that helpful if it’s not packaged with empiricism, an epistemological way of acquiring knowledge.

To even get off the ground analyzing data, we need theory of measurement. What should we track, given limited engineering resources and raising storage costs? Claiming we should track, say payments and logins, at the exclusion of audio volume, implies a cost-benefit value ranking. Why are payment and login more value to track? The theory is that understanding payments and logins will unlock more insight then volume as volume plays a less significant role in the app. Is this true? Hopefully, the institution has the intuition or previously collected to knowledge to make an educated guess. Firms have discovered that refining this knowledge can make their bets more likely to succeed. As it so happens the West has created the best knowledge refinement process in the history of humankind: the scientific method.

Data or more broadly, empiricism, is a key part of the scientific method as it expands the sample size of a test beyond antidotal evidence. Doing this at scale, as well as the methodology of running true experiments, A/B tests, means that knowledge is more valid (less likely the result of antidotal evidence) and stable. Firms can now learn.

Arguing to be data-driven or informed misplaces the value in the supply chain. We need to more explicit in this endeavor – it’s not about data, it’s about science.