The PC games pioneer bares all on Valve's unique dev culture, and shares his grand plan for the future.
[This Q&A comes as part of Develop’s package of five interviews with Valve Software. An index of each interview can be found here.]
Gabe Newell - Co-founder and company president
So, honestly, is it true that everyone at Valve has wheels on their desks and can change what floor they work at whenever they like?
Yeah, you know why right?
Because people get annoying?
[Laughs] Actually it’s quite relevant to what we’re talking about.
I’ve run lots of different kinds of organisations, and each one needs to be suited to the task they are performing. Hierarchical organisations are pretty good if your business operates on repeatability and measurability, and in fact they cope fairly well with the loss of personnel.
A lot of these concepts originated in the military, in fact, which of course often has to deal with a loss of personnel. The holes in a hierarchical set-up can be filled by new people and the business goes on.
Yeah. In fact, factories were reorganised around the military concepts in the early twentieth century.
But if you’re trying to invent things, or do novel things, a really strong hierarchical organisation can get in the way of that.
The point being that, if you’re constantly having to change, rigid notions of organisation get in the way. If you look at how quickly the video game environment is changing, what works really well in one generation becomes pretty irrelevant in the next. You go from sprites to polygons. From 256-colour 64x64 bitmaps to shaded polygonal models. Game development studios have to constantly keep reinventing themselves, processes have to change over and over.
That’s why Valve is organised to find people that are cross-disciplinary. [Valve developer] Ken Birdwell is a fine-artist that taught himself to program. He did the skeletal animations for our games starting all the way back from Half-Life 1.
Do you remember the tentacle in Half-Life? Well, Ken was able to edit the level, change the tentacle model, and change the code, in order to make that all work.
If he’d have just been a modeller, or animator, or environment worker, or just a programmer, he would only know one way to solve a problem.
So, essentally you're saying that Valve needs autonomous developers.
Yep. We don’t really have titles here; people decide for themselves what their role should be. People self-organise here.
That’s not out of some nutty thinking – it is the only way we can properly solve the kind of problems that we need to solve, and do it quickly.
So, to answer your question, that’s why we put wheels on desks.
People say they have to go work on this different problem now, because nobody else is and we need to get it done in order to get other things done. So they pack up their desks, and move to their relevant teams.
That makes you a more fluid company, I would imagine. It would allow you, in theory, to quickly adjust to the changes happening in the industry. Is that the wider plan?
Yeah, we think we have to be this way. Right now it feels to us that the games industry is changing faster than it ever has since we started Valve in 1996. As much as we’ve tried to be flexible and adaptable in the past, right now it’s more important than ever. Things we were successful at in the past don’t matter a whole lot in predicting how well we are going to do in the future.
I have a theory that Valve has entered its third phase. In the early years, Valve made isolated boxed product, and later it went on to make net-integrated games. These days, with all the different media Valve is working on – and the way it approaches this – I feel the company has become an online ‘entertainment’ company.
Well, the terminology we use a lot is ‘entertainment as a service’.
We constantly ask ourselves the question, how can we make our fans happy every day, and what are the different ways we can do that.
The internet enables a really efficient delivery of… stuff, to your customers. But we have to do this because audiences are really mobile, no one is going to sit around for five years waiting for you to get something done, they have too many other things to do these days.
When you look at the Team Fortress 2 shorts, we don’t think of it as a games company making something other than games, we think of it in terms of the audience and whether they respond well to something we’ve done.
So in the case of the Team Fortress 2 shorts, people responded really well, so we should do more of them.
We really don’t know how to make decisions without constant feedback from our customers. The faster we sort of iterate on that, the happier our customers seem to be.
Half-Life 2, 2004
The interesting thing is, that approach leaves you with so many avenues to go down. And as stretched as you are with resources, I have no idea why you don’t hire 50 new people to get more done.
Well, we’ll hire anyone who walks through the door who can pass our review process. The problem is very few people can do that successfully.
The cost of lowering those standards is huge. If any of your readers are interested in the philosophical underpinnings of some of our decisions, The Mythical Man Month is a book that is a pretty good start about what we’re optimising for.
In the games space, or any rapidly changing space, the fewer people that can get something done, the better it will get done.
Additional people on a project have their own negative externalities associated with them, and so you’re always better off getting five people to do something than fifty people, as long as the five can do it. The key is finding those five right people, and that’s an incredibly hard process.
But people we want to work with, we’ll try hiring them for years.
Like Doug church.
Doug Church, Michael Abrash [renowned programming author, Id Software, RAD Game Tools], Mike Sartain [Bungie, RAD Game Tools], Scott Ludwig [14-year Microsoft veteran, contributor of the first Windows OSes].
The first time I worked with Scott was in 1983, and he’s one of the strongest programmers I’ve ever met. He recently started working here, and he’s one of those people like Michael Abrash and Doug Church that makes everyone around them work better.
You know, it’s not a case of anyone ever needing to catch up on Michael Abrash to make sure he’s not wandering into the weeds. You catch up on him to find out he’s working on something amazing and magical – which would usually be a side-affect of him solving a problem that was annoying him [laughs].
The point being, when you run a company composed of people like that, good decisions and good ideas happen all the time. That’s decisions made without the need to consult upwards or direct downwards.
It’s completely different to the way the rest of the industry is going, if you ask me. If you look at Ubisoft Montreal, or Ubisoft Toronto, these are 1,000 person-capacity studios that can knock out a game in twelve months.
The only problem with that is that they’ll be able to knock out the same thing over and over again, not something that adapts to the changes in the industry.
And I haven’t seen any evidence that the rate of change in the industry is decreasing. I think it’s increasing. So what you’ll end up finding is those thousand people are the enemy of your next project. They may be able to bang out your current one, but they’ll get in the way of doing the next one.
But, seriously, if fifty awesome people knocked on the door, we’d hire them all. We don’t hire to specific positions, we hire to standards.
It's nevertheless dificult to understand how Valve works in practical terms.
Usually we say it takes people about six months to get the hang of it. One of the first things we do when new hires come in is we pick a project that’s in the sweet spot of their competency.
We pick something they’ve done before and have done well, and then we try and make sure they can deliver that to customers. In other words, we don’t say they’re working on project X and it will ship in three years. We try to put them on something where they try to create value for customers, and do so in the near-term, so they can watch the whole process and iterate on it.
The aim is so they can say, ‘oh my god, I’m here at my desk, I’ve made a change to a game, and 30 to 40 million people will see it in a few hours’.
It’s a very different environment than normal. A lot of the time our new hires ask us, how do decisions get made? How do you pick which projects to work on? People think there must be some secret cabal somewhere in the company that’s making all these decisions. But after about six months in they get it.
Half-Life 2 Episode 2, 2007
What about when you’re in full production though? What about when you have six months before a project is released?
And so people pick up their shovels and start digging. Since everybody gets used to this idea and no one tells them what to do, everybody gets used to this idea of picking what’s best for them to work on. And it’s amazing how much more productive people can be when they’re setting their own agenda.
Well I think we’re pretty productive. If you look over the years of what we’ve produced, I think it’s evidence that this approach works.
And, we have almost no turnover at all. Once people get here they think it’s awesome. They want to stay here [laughs].
One thing to think about is that, when I was at Microsoft, after I left I was very lucky. When I joined Microsoft they were the third-largest software company on the east side of Lake Washington!
When I left, I could retire or do something I wanted to. Not because of something I had created, but because I was in the right place at the right time with a bunch of other really fortunate ‘lottery winners’.
Then I realised that, since I could do anything that I wanted to, I decided I wanted to work with other really smart, motivated, socially-orientated people to create product that would affect millions of other people. To me, that was the most fun I could have. That was way more fun than getting on a sailing boat and going round the world. I would’ve made it half way to Hawaii, cut my wrists and thrown myself to the sharks.
To me, this is what I would do if given the choice to do anything. For a lot of the people here that’s true as well.
And it seems like Valve is a studio that is trying to remove money from the equation. In that, people will have enough money to not worry about it, and can focus their attention on other things.
I think it’s obvious, everybody here could get a job in ten minutes. People like [someone walks into the door, it’s Jay Stelly, holding two coffees], people like Jay! Who brings me coffee.
Jay is the most senior engineer at the company [laughs]. And he saw that I didn’t have coffee! There are probably at least five publishers he could call and get a $30 million project in an hour if he wanted to.
The reason he’s here isn’t because he has no other option, he’s here because he can work with the best people we have. It’s here at Valve where his talent can make the biggest impact and he doesn’t have to waste time rolling his eyes at stupid things the marketing and management teams are doing.
That’s not by accident. That’s the whole point at Valve. That freedom he has to do the right work is a consequence of our whole philosophy of how a games studio should work. If we were in the business of making LCD screens, we’d have completely different management practices.
Team Fortress 2, 2007
In the last three years there has been massive shifts in social and mobile gaming, they have both become huge sectors for the industry. Is this something you would like to be more involved with?
Well, the thing we’re trying to think about is where these two sectors are going. We look at mobiles as a very important form factor, much in the same way living room devices are important form factors.
We’re comfortable in delivering PC stuff to the console side, and I think Portal 2 will cement that fact.
On mobile, however, we’ve done nothing. Especially compared to the guys at Epic Games. I was talking with Tim Sweeney and Cliff Bleszinski recently and I told them how super impressed I was with what they’ve accomplished on mobile.
In fact, I was actually surprised with how excellent Infinity Blade was. I was playing Infinity Blade and I thought, wow, I had no idea you could get that out of an iPad.
The thing is, clearly the mobile platform is coming along. The thing I hate about it is we have to live with the CPU and GPU constraints that we’ve not had to deal with for a while.
Actually, I also hate the delivery through cellphone technologies is so outrageously priced. It’s hard to deliver a large game when you’re paying $12 per gigabyte delivered in the US – right now in the console space we don’t have to think about size. But now in the mobile space, customers want to pay $2.99 and you’re thinking [laughs] well so much for our fifty gigabyte game!
[Laughs] “That’ll be, um, $600 please”.
Yet, the thing I like about the mobile platforms is that they start with the presumption that they’re already connected. They are ahead of the consoles in that sense.
We love Nintendo to death, but they still really struggle to think about what the Wii means to be a connected device, and what their games are in terms of a connected experience.
What are your views on the social space?
Well, I think in that whole space there are so many different things to draw on.
First you have the traditional small multiplayer group that Id software started back with Doom. You have the MMO model. You have the Farmville model.
At Valve we’re looking at each of these as different pieces of a bigger puzzle. We’re trying to work out the strengths and weaknesses of each and what we can learn from them.
I mean, it’s amazing what the Zynga guys have done, and some of it seems so un-examined. Just their ability to reach 84 million users in a month!? Just compare that to the total size of an entire console platform over its lifetime. Clearly there’s something valuable there, there’s something to learn in that fact.
Is Facebook a viable platform for Valve?
We tend to think of things in terms of our customers, so if our customer is a heavy Facebook user, we have to think about how Team Fortress 2 or Portal 2 can be better for that customer.
So when we look at Facebook we see a collection of services, and a lot of our customers there, so what’s the useful thing we could do for them.
The point is, too many of our customers use Facebook for us to ignore it. The same way too many of our customers have iPhones and Android phones for us to ignore it.
Your audience is already there.
Right. And that’s how we looked at it when Steam first started. We honestly felt, with Steam, we were actually really late in responding to the way our customers were using the internet.
Now that I look at mobile and social, I think we’re late again. We just need those magical fifty people to walk in the door we’ll all be set!
Left 4 Dead, 2008
After Half-Life 2 you had openly discussed the health and wellbeing of your staff, and how you were going to attack this problem by adopting a less intense episodic production model. During that five-year Half-Life 2 project, did you personally feel a sense of impact you were having on your teams’ health?
Oh, absolutely. I’ve become obsessed with this issue now. In fact that’s why we’re all going to Hawaii in a week.
We do it every year. Usually I think it’s more a benefit to the families of our teams – I mean some people here you literally have to drag away from their desks.
But the families around our employees usually love it, because they get to see other people’s families, they get to see why their spouse or parent is so fired up for what they’re doing at work, and the kids have a huge sense of pride.
Do you think the episodic model works though?
I think that we accelerated the model and shortened development cycles with it. If you look at Team Fortress 2, that’s what we now think is the best model for what we’ve been doing. Our updates and release model on that keeps on getting shorter and shorter.
We went through the episodes phase, and now we’re going towards shorter and even shorter cycles.
Left 4 Dead is starting to approach the Team Fortress 2 cycle. Portal 2? We’ll have to see how much our customers want us to push in that direction. In general, our approach is to come into work and ask ‘what can we do for our customers today’?
We get a huge amount of value in releasing things. Every decision you see our Team Fortress 2 team make is a direct result of feedback they’re getting from customers. Everything our team does is a result of tests they’ve done on the last two or three releases. Because it’s information from the last few updates that tells our team what to do next.
Reminds me of Zynga, to be honest: Releasing a product and consistently updating it.
If you talk to some of the Korean developers, they actually make fun of us for taking so long to do updates. They say that, until we release updates every single day, we’re missing a huge amount of value. I think there’s a lot of validity to that perspective.
There are also a lot of challenges to that, of course, like how do we safely make large-scale changes to our games within their own constraints? How do you become a novel when you’re a newspaper?
These are things we have to figure out, but we sort of amortise the risk by working on different frequencies for different projects.
Team Fortress 2 is the fastest frequency we work on with comparatively fast updates. Er, Half-Life is apparently the slowest! [Laughs] Although, from the outside world, we have no evidence that Half-Life is working on any frequency at all. [Laughs]
You seem happier than ever. I’ve noticed this in a lot of the interviews you’ve done recently. You’re laughing a lot, making more jokes, taking things less seriously. It’s just a far cry from the last few months of Half-Life 2 when you seemed visibly stressed.
It’s hard for me to look back at how disgruntled I was before Half-Life 2 shipped. I reckon everyone here’s really happy at the moment. We certainly feel pretty great right now... knock on wood… pride cometh before the fall and all that. Now a meteor will probably strike us in the middle of this interview. [Laughs]
I suppose this is the end of Valve’s big, five year projects.
I’m not entirely sure. You want to distribute your choices. Right now there’s a bunch of pressures to have shorter and shorter development cycles. But that could change.
I’d have to find a reason for it to change, but it could. I don’t want to be caught completely off-guard and overly invested in one area. I think you’ll still see projects from us that are huge in scale, simply because we have the ability to do that.
Am I right in thinking there have been two attempts in changing the way Valve works since Half-Life 2? The first being the episodic model, the latest being the ‘entertainment as a service model’?
Yeah we’re thinking of this all now as entertainment as a service. We’re now fully focused on asking how we can take advantage of being constantly and fully connected to our customers. We now work from data we get back from our customers, reading into what they actually do.
For me, ‘entertainment as a service’ is a clear distillation of the episodic content.
Is Source undergoing a major update?
There are lot of advantages on iterating on a mature and stable and shipped codebase, as opposed to starting over again.
I think, when you see a game like Dota 2, you’ll see how developers can get a lot more out of Source than most companies can get from a scratch-built engine. I think that incremental approach to Steam has worked really well for us.
Does that mean we’ll reach some architectural tipping-point where we’ll need to change? No. I mean, if Larrabee [Intel’s promising GPU that was cancelled as a consumer device] had shipped that would have probably necessitated some fairly dramatic changes in order to take advantage of it.
But, so far we’ve been able to keep the engine moving ahead, robustly. I mean, I think it looks great.
What surprises me is how few other studios use source.
Yeah, I think a big reason for that is not having a PlayStation 3 version. When you look back at big game engine decisions studios were making a few years back, we didn’t even have an Xbox 360 or PS3 edition of Steam.
I think going multi-platform was a pretty critical decision for a lot of studios. Also, Source just wasn’t a high priority for us. We’re really happy if another studio wants to use our engine, but we’re not going to go out there and try and muscle in on what Epic Games does.
A few people have used our engine, and I think a few more will find it useful now that we have a PS3 edition.
Portal 2, 2011
Now you have the PS3, Xbox 360, PC and Mac support, is Source a business model you want to pursue more aggressively?
Y’know, we’re happy if people want to use our tools. We’re also super happy if people want to use Unreal Engine. We’ve worked hard with the guys at Epic Games to integrate Steamworks into Unreal Engine, which we think will be a great solution.
Our philosophy is always about creating the best value for our customers, but also our partners, and right now I think there’s more value for us to pursue things like the microtransaction part of Steamworks.
I think if we’d take the microtransaction model away, and instead push harder on getting studios to sign up to Source, I think we wouldn’t be using our time nearly as efficiently. Y’know, if people want to create a mobile game today, call Tim [Sweeney, Epic Games].
We’re more looking into what Valve can do with payment services than engine services.
The industry has this broken model, which is one price for everyone. That’s actually a bug, and it’s something that we want to solve through our philosophy of how we create entertainment products.
What you really want to do is create the optimal pricing service for each customer and see what’s best for them. We need to give customers, all of them, a robust set of options regarding how they pay for their content.
An example is – and this is something as an industry we should be doing better – is charging customers based on how much fun they are to play with. Some people, when they join a server, a ton of people will run with them. Other people, when they join a server, will cause others to leave. We should have a way of capturing that. We should have a way of rewarding the people who are good for our community.
So, in practice, a really likable person in our community should get Dota 2 for free, because of past behaviour in Team Fortress 2. Now, a real jerk that annoys everyone, they can still play, but a game is full price and they have to pay an extra hundred dollars if they want voice.
That’s just one example. Another is how much people want to pay for items. Some people are happy paying a dollar. They’ll pay a dollar over and over and over again, others want to be different, others want to run servers and create mods.
Each one of these people should represent a different monetisation scheme for the community as a whole.
So, one thing we do have are these high value customers. We’ve now started connecting their Steam account to their PayPal account, and now these people aren’t just paying for games, they’re making money from them. And it’s not just a little bit of money, it’s $20,000 per week some people were making.
Their cost for Team Fortress 2 is negative $20,000 per week. You’re never going to see that in a retail store. [Laughs]
That’s the best thing in the long run for the community. Our challenge is to knit all the positives and negatives of our community together so that everyone who is one Steam is adequately tailored for – and that doesn’t at all look like a RRP on a box. It’s people who make hats get paid. People who are really popular play for less, or free.