Industry Lifecycles

Here we go again – “________ are dead! ______ is the new hot thing in videogames!” Whether it’s Indiepocalypses or E.T. carts in Alamogordo, haven’t we been here before? Why yes, yes we have. Some things march forward – like costs – and some spiral around the same familiar patterns. In this talk Raph will describe general patterns of platform shifts in the game industry, the inevitable logic of maturing markets, and how that impacts what the audiences look for.

This talk was presented at Casual Connect in LA on January 16th 2008. It is based on over 250 data points on actual game costs, adjusted for inflation, compared to the typical lifetime earning per customer for that game, and compared to the actual install size including streamed down data. It is a follow-on to the talk “Moore’s Wall” from 2005.

Slides

 

Video

Article

I wrote an article based on this same data, which was reprinted on GamesBeat and Gamasutra, and covered by numerous outlets. You can find it here.

PDF

Transcript

RAPH KOSTER: I swear, even though this is mostly business data slides, that I will have a slide of designer-y advice at the end. Any of you designers out there—I see some that I know—good, good. You’ll be reassured. Some artsy-fartsy stuff is in fact on the menu today.

So the problem is, we as creators, as artists, as developers, work inside the context of business people. They tell us, “This is how much money you get.” They tell us, “This is how much time you get.” They tell us, “No, you have to sell loot boxes.” They tell us, “No, it has to be purple this time, or pink this time, or whatever, because that’s the fad.” And they tell us, “It’s got to make back this much money or we’re canceling your series, and you need to go find a new job.”

We know that. So understanding business, for us, is a survival tactic. If we want to be successful creatives, we’ve got to master it. We’ve got to understand it.

It helps that, given the track record of the game industry, those of you who have been around a while may agree that a lot of the business people in the game industry haven’t mastered it either. So we’re on even footing there.

Clear back in 2005 I started looking at the question of, “Hey, how is it that this business works? Why is it that new things come along and new games start succeeding?” It didn’t take much to go looking in some business textbooks and arrive at some pretty standard lifecycles and all of that kind of stuff.

We all know this stuff. A new platform comes along, and eventually we get winners, and then eventually a new platform comes along and some of the old dinos die. It’s pretty familiar. We’ve had that happen a whole bunch of times over the course of the industry, with different-sized platforms, some of which we don’t necessarily think of as platforms.

I’ve come to think of them all as consoles. Facebook is a console. Flash is a console. iOS is a console. They all pretty much follow the same cycles that we’ve always seen. It doesn’t matter whether we’re talking MMOs, or phones, or Flash, or Facebook, or PHP web games.

There are almost always the same characteristics. They’re worse at whatever kind of game is already hot. They’re worse at doing the current hot thing, and they’re better at something new.

Then what happens? It’s cheap to get into, relatively speaking. It’s unproven, so there’s no EA around. That’s good, right? Less competition. It offers new controls, new capabilities, new constraints, whatever. That means the old guys can’t compete effectively, and those of us who are nimble little mammal-rats can get in there quickly and start learning all the new stuff before they figure out that they can spend us into oblivion.

That’s great, because they do eventually spend us into oblivion. That is, in fact, what happens. They spend us into oblivion.

We expect the early entrants to come in, and they make a pile of money, and as they become big guys, they use that money to kill us. They use that money to kill all the small guys. Sometimes literally, as in cases where, for example, Zynga intentionally outbid everybody else on the Facebook ad platform in order to prevent new entrants from gaining traction.

That’s the game.

The result is, these are all companies that have pulled this off at one platform transition or another. Zynga managed to do it with Facebook. Rovio managed to do it with mobile. PopCap did it with the web game. Sony Online did it with the MMO. EA did it clear back in the original old-school days of the early ’80s. They managed to pull it off with PC gaming.

Sometimes whoever succeeded last time falls flat on their face when the next turn in the road happens. We saw SOE basically stumble and fall as an incumbent. We saw PopCap ironically get swallowed by EA when the nature of casual gaming changed.

Sometimes incumbents do make the turn. Zynga, we were worried about you for a while there, but two wheels off the road, back on, you kept going. Awesome.

But it can cost a lot of money for those big incumbents, and sometimes what they have to do is buy a bunch of the mammals and incorporate that into their DNA.

Then this new set of winners says nobody else is allowed in now. They raise costs. They focus on brand rather than on game mechanics, because brand is defensible and game mechanics aren’t. IP lawyers, you should fix that someday.

The midsize guys either have to spend to compete with the EAs and Zyngas and Rovios and Supercells, which they can’t, so they get bought, or they shrink and they die.

The small shops occasionally get Flappy Bird, and that’s great because that generates news items that encourage more victims—I mean new entrants—to come into the market. But it’s basically a lottery.

We’ve seen this. The funny thing is that, depending when you first started video gaming—and by the way, some of you right now are hearing a soundtrack in your head because you can’t help it, and I know exactly how old you are. The rest of you will have to go find that game.

If you came up during those glorious periods of a brand-new platform, you love gameplay systems and gameplay mechanics, because you grew up a neophile. You grew up somebody who loves seeing new systems develop and crazy new game-design ideas happen.

If you came up during mature periods, you love stories and art and brands and all of that awesome stuff that today, because we’ve been in a mature period for a while, we tend to associate with whatever is triple-A.

That was fine as an explanation for what it did. I mean, I just wasted five minutes telling you something you all kind of knew. The problem is that doesn’t tell us anything about where we’re going. We can’t validate it. It doesn’t really tell us what comes next.

To answer that, we would actually have to have real data. We would have to gather a statistically significant sample of highly confidential information that no publisher would ever release, so that we could actually assess the health of the market and where the trajectory is going to go.

Surely nobody could actually manage to acquire that unless they had a whole bunch of friends who were willing to break NDAs and talk to you off-books.

So I did.

This is a sampling of 260 to 270 games dating back to the mid-’80s, with the raw development cost in inflation-adjusted dollars. These are the real numbers: how much it cost to build the games, marketing excluded, in 2017 dollars. It is across all platforms.

That’s kind of hard to read this way. You might notice: what is that curve, really? So let’s change it up and look at it on a log scale. What does a log scale mean? It means an order of magnitude, ten-Xing on every line here.

This is a curve to the moon. This is our development costs today.

Hey, look, Flash was born. Hey, look, mobile games were born and Facebook games were born. Boo, they’re all moving up to almost the same slope, sometimes faster. Hmm. That’s weird.

Well, that curve doesn’t look good, considering it’s a highly sloped exponential curve. Wow. That line, hitting 2020, is about $200 million average cost.

Observation one: costs 10X every ten years. Raw development cost 10Xes, and it doesn’t matter what platform you’re on.

Now, the kinds of data people tend to share: a bunch of this data set is private, and a bunch of it came from investor reports and things like that. What numbers tend to get shared on budgets? The really high ones, because they make good headlines. So maybe there’s a bit of an over-report bias.

There’s probably also an under-report bias on the really low data points, because indies conversely brag about how cheaply they made that.

If you take this inflation-adjusted curve and look at the average development cost, you can actually see where we have not enough data points prior to 1995 in the data set. Also, it turns out people don’t want to share things from the last two years, but they were willing to slip me more numbers the further back we got. So there’s definitely some wobbliness in the numbers through there.

Also, reporting on free-to-play games: it’s awfully hard to get a lot of numbers on those.

The funny thing is, if you look for graphs like this online right now, what you will tend to find is really nice-looking step-by-step graphs that show platform transitions. Yet that line was pretty smooth, which is interesting. You’d expect it to step. You’d expect the widespread use of Flash, or Unity, or other engines to cause stepwise functions in the efficiency of development, and we didn’t see that kind of thing.

So what is driving these numbers? I went looking, and it turns out there is one answer: raw bytes in the install. It is the simplest proxy. How big a game did you make in absolutely raw bytes?

This means for a mobile game, I went and checked, okay, it is this big when you initially download it, and then it streams a bunch of crap to you. How big is it after all of that? If it’s a PC game, it is not the installation requirement number needed, it’s what’s left after you’re done. So it leaves out all of the temporary copy space.

Observation two: bytes, meaning content, are what’s driving this insane cost rise. They’re probably underreported, because for example there was a stepwise function at a version of the Xbox where compression went in, which shaved thirty percent off every single hard drive install on the Xbox, and you can’t even see it on the chart.

So our cost per byte works out at—anybody want to guess? Are we getting better at it?

We are, thank God. However, this is also on an exponential curve.

The biggest oddity about this is that in the space up through 2005, we got six times more efficient at generating the 122 times more bytes that we needed to create. Since then, it’s actually been roughly flat, regardless of whether you work in indie, in mobile, in casual, or in triple-A, because salaries are salaries.

It’s been kind of flat, roughly, since Unity and Unreal 3 were launched. That might be because they standardized the process, and radical improvement in content generation ceased because the tools were so good we started using them as they were.

That is actually bad. We want the curve of cost per byte to keep dropping. We want it to drop faster if possible, so that we get higher ROI from our spending.

That was a troubling result for me, looking at these numbers. So then I decided to go see: what was the price of that game at the time that it came out? Let me adjust that for inflation and figure out how much a player pays for each byte we give them.

It’s no surprise, hopefully, to all of you that the actual purchasing power of players has plummeted. A game from the 1980s today would be at approximately $120 at retail. But because of inflation adjustment, players are now paying half that, literally half that, for a triple-A game.

It turns out this is also on a log scale. Players are paying a disturbingly low number here.

What are the things that stick out? Where players pay more per byte, the dots that were up in this range, almost all of them are MMOs or ongoing services, where the lifetime value of the player involved a box purchase plus months worth of subscription revenue. It was not unusual to hit $150 to $180 for every single user who touched the game, on average. So an ARPU of $180. That’s what held the curve up there.

For everybody else, per byte is now down at around thirty-six cents. So if you ever need an illustration of race-to-zero pricing on Steam, on mobile, this is it. This is what is happening. The cost per byte is plummeting, and it still costs us the same amount to make.

Now, prices have actually fallen more than this shows, because this data set underrepresents free-to-play. Of course, free-to-play also shouldn’t register at zero, but getting good numbers on LTV out of free-to-play companies is noticeably challenging.

There is a lack of mobile games in the data set. There is no way to really find out Steam sale average real pricing, so all of the triple-A games are actually over-reporting pricing, because I left them at the sixty-dollar mark, and we know that’s not accurate. How many of you have hundreds of games in your Steam library?

Also, service games tend to stream bytes where possible. Those are actually in this count. The streamed bytes are in this data. Mobile install size includes the stream size.

The previous slide and this one maybe cancel each other out. Hmm. I don’t know. Fact of the matter is, the trend line is so strong, I’m not sure any of that wouldn’t just be noise in the data.

Service games have ongoing revenue, which is great. Is that enough to cause a change in this trend line in the data? I’m not sure it is.

Here’s the fundamental thing that trumps all talk of circular life cycles in the business: the business is a spiral, not a circle. We are always moving forward on a technological advancement track: Moore’s Law, all of that. We have to keep filling that larger RAM, those nicer video cards.

The fact of the matter is that these two lines are diverging. This here is the development cost per megabyte, and you can see it’s pretty flat since those engines became dominant. This is the price paid per megabyte, down to thirty-six cents per meg paid by a player.

The further apart these lines get, the smaller margins get. More importantly, these are regular enough that we can forecast. If the slopes on these lines don’t match, what does that end up meaning?

Well, not to bring you down, but it could be bad news for all of us.

First observation: the average game will cost zero to buy in ten years. Byte value will be zero, kind of where music is today, lest you think that it’s implausible. But it will still cost $3,205 to make that megabyte.

Forecast number two: triple-A games are going to cost, on average, $200 million by the early 2020s, unless they start making different kinds of games.

Now, tentpoles will be pulling that up. Maybe they stop making the equivalent of Destiny, and maybe this curve bends. That’s entirely possible, because after all, that would mean they’re shipping a terabyte. So it might not be shippable. It might not be storable. But yeah, I don’t know. Terabyte drives are getting cheaper.

It might mean that triple-A has to shift, change the nature of the kinds of games they make towards things that require fewer bytes, and I think we’ve already seen a bunch of that shift starting to happen today.

Now, I say it’s speculative, but you want to know why there are loot boxes? This is why there are loot boxes. Because we have to find ways to push up this revenue: upsell revenue of any sort. Loot boxes, free-to-play, subscriptions, tiers, membership clubs, Netflix, it doesn’t matter what it is as long as it’s an upsell over the base price.

However, we’re already doing all of those things, and the curve keeps marching on, because we do it via content. Why do we do it via content? Because we can sell content. We don’t sell systems. We sell content. We sell digital stickers, and now we’re going to put them on the blockchain so we can sell them better. But they’re still content. They’re still bytes to generate, and that doesn’t actually solve the problem.

We’ve started drifting towards more runtime procedural generation. Think No Man’s Sky and so on. Of course, it’s hard to talk to a mobile developer these days without bringing up the question, “So you’re really kind of making an MMO?”

There’s a reason community, user-generated content, esports are booming. In part, because it demands high skill ceilings, more gameplay time per byte, more retention, and changes the slope of the curve.

Even with all of those things, I worry about the sustainability of this curve.

Why have we gotten away with it all of this time? Because the audience kept growing. But we crossed the fifty-percent mark of all North American people being game players in 2009. We’re actually running out, on a global scale, of wealthy people to turn into gamers.

Audience expansion is capping. It could be that maybe the rest of the world, their standard of living and the currencies all skyrocket, and that’s great, and they can afford sixty-dollar games. Or U.S. prices go up. But that looks to me like it’s just going to happen because otherwise the businesses themselves won’t be able to cope.

Everything I just told you isn’t grim enough, because I left out marketing. The more mature the market, and the more you compete on brand, and the more you try to reach extended addressable market, the more marketing budget you need.

In triple-A, marketing budgets are as much as that development cost again. In mobile, they’re sometimes ten times the development cost budgeted for UA.

So what saves us historically? That platform shift, right?

VR, unfortunately, wasn’t it. It turned out VR was just as expensive, if not more so. It was harder to develop for, higher required frame rate, higher technological skills required. Maybe some of the cool new phone AR stuff. Maybe it will be new forms of streaming. Maybe. I don’t know.

But I am going to be looking out for it, because that’s where all the mammals are going to jump ship to. VCs, keep your eyes out.

What does this mean for you, those of you who raised your hands and said, “Oh yes, I’m a designer”? We have gotten used historically to thinking of these cycles as, “Oh, this segment is now mature. Let’s move to the new segment.” What these curves are saying is that the segments are kind of all maturing, because all of their costs are rising in tandem.

What can you do if you are in a mature market and you are thinking in terms of your competitiveness as an individual designer?

First, get good at systems, not just content generation. Why? Because systems drive player longevity. Player longevity and retention drive lifetime revenue. It’s inescapable. So less cutscene, more game.

Brand-build like crazy. Goodbye, abstract puzzle games. They won’t be viable, because you need branding to stick out in a market like this.

Static content is your enemy. You need to start thinking of how to avoid static content. Get comfortable with those techniques today. There’s all kinds of fascinating stuff happening in the academic market. They are procedurally generating Zelda at this point, the original Zelda. That’s how far academics have gotten with this. Industry folks need to start reading those cool papers.

Don’t even bother building a game or conceptualizing your next game unless you can think of how it will be a service business in some form or fashion, because what ends up really driving those kinds of long-term revenues is going to be community. That’s the thing that will actually make a difference.

By the way, if you have not been working in a service business, I estimate it takes about three to five years of learning curve to learn how to do it. So start now.

MMO is basically the same idea: service business, community, retention, UGC, deep systems.

On a personal level, speaking just to those of you who are creatives in here, this crunch hits the grunt developer worse than it hits the publisher. If you work on the business side, great, because the work you make goes into creating an IP for the IP owner, not for you. The IP owner gets residuals forever. They get brand extension.

It’s time for developers to start thinking about how they can get residuals for the concepts they create, because otherwise this picture looks really, really bad for creatives.

Now, I could be wrong. So send me more data. You can send it there. I need name, cost, year, platform, total bytes, if you know it. The data set will only get better.

In the meantime, winter is coming. Invest in parkas. That’s my advice.

As a designer, think about the ways in which you can learn how to keep the home fires burning. Thank you.

AUDIENCE MEMBER: Do you think AI might change the authoring equation a little?

RAPH KOSTER: AI—artificial intelligence—might change the authoring equation? Yeah. AI is another form of procedural content generation, so yes.

Here’s the thing. You know Miller’s Law is that software is a gas that expands to fill its container. Not too long ago, I was having a conversation with a mobile developer who makes a match-three RPG, and they were telling me about how many EC2 instances they needed, and how critically important it was that there be absolute server verification of everything in their match-three game on the server, and how incredibly expensive it was to run all of that stuff in the cloud.

I went, oh my God. We were running UO servers on Pentium somethings, and I’m pretty sure one UO server could have run ten thousand match-three RPGs in it.

We have gotten very comfortable with static content output, with simple state machines, no AI, no proc gen. We basically don’t really use the computers very much in what we’re making. We use a lot of what you make, and that’s actually part of the problem, referencing video cards. We are spending more on pixels than we are on things like AI and the rest.

Almost every other route that I listed here is a minor one that bends the curve. Getting smarter about using our computing resources is probably the only thing where you could literally make an order-of-magnitude difference tomorrow, if you just went and threw away your server and wrote a new one. That buys you ten years. That’s better than nothing. So yeah, AI, absolutely.

AUDIENCE MEMBER: Raph, when we talked about this, you covered most of the three ways out of the box, but I think you should give a warning about just how difficult UGC is, and about the debt. You only covered community in terms of the wrapper.

RAPH KOSTER: Yes, I felt pressed for time. There are ways out. You can pursue UGC. That’s a way to avoid creating bytes. You can pursue procedural generation. That’s a way to avoid creating bytes. AI is a way to avoid creating bytes. There are multiple doors out of the trap.

All of them are actually really hard. Although you can find historical examples of each of them working successfully, you will actually be hard-pressed to find sustained success historically on them.

We can all point at Minecraft as being extraordinarily successful at leveraging UGC and community in order to sidestep this problem, but it is sui generis. It’s the only Minecraft we got.

That is partly why I want to ring the warning bell now. They’re all actually extraordinarily difficult solutions that we as an industry have not been training ourselves to execute on. We have trained ourselves to execute on polish, not on these other approaches.

So yes, the point that all of these are extraordinarily difficult is very valid, and people need to start working on them now to build a competency. Even then, I think it’s no guarantee.

AUDIENCE MEMBER: Sorry, is it that MMO has worked, UGC generally hasn’t worked?

RAPH KOSTER: The MMO body of work exists and has worked. UGC generally hasn’t worked. So some are even more difficult than others, and you should put your people towards the ones that have worked.

Community is probably the easiest, and those of you who have worked in community know that when I say it’s the easiest, that should give you an idea of the relative difficulty, because community is really, really hard. It’s probably the easiest solution.

AUDIENCE MEMBER: [Question about where the data came from.]

RAPH KOSTER: I got the numbers first from a substantial number of investor reports and contemporaneous press reports. In all cases, if marketing numbers were included and signaled as such, I stripped them out. Then they were all inflation-adjusted. If the numbers were given in euros, yen, zlotys, or whatever, I converted to U.S. dollars baselined on the exchange rate of the year in which the game was launched, and then inflation-adjusted the dollar forward.

Everything else came from private discussions with developers.

AUDIENCE MEMBER: With your experience and research, do you think there’s any viability in smaller developers staying small and focusing on niche targets, and never actually growing? Keeping their teams small and servicing such a small audience that large overs would never take it?

RAPH KOSTER: There is. It’s almost like—there is a recipe that works there. It is actually viable today in the mobile market and in other low-end markets, and that is: acquire a pile of money. Which could be you eating ramen. That version of the pile of money.

You have enough of a pile of money that lets you make ten, twenty things that are very tightly scoped. You basically put money in the slot machine. Do everything you can—it’s not that you can weight this bet, but weight it with polish, with brand-building, with everything else, absolutely every unfair advantage you can, and hope that you get one of those to pop.

Once you have one, you do it again, and you begin to cross-market, and then you start trying to extend community. That is actually working for multiple companies in the mobile space today. Many of them are fairly small, and they plan to stay that way.

The real trap on some of these is actually doing okay. If you get to the point where what you’re trying to do is sustain a service business and you can’t afford to go make another game, then you’re screwed, because then what you need to do is shoot your current one in the head, and that effectively shoots the community because it will never trust you again.

So it can be done. You have to constrain your ambitions, and that precludes a whole bunch of funding sources. Anybody who wants an exit, for example, is no longer a viable funding source if that’s the model you’re going to pursue.

One last ten. Thank you.

[Applause.]