Valve on Source and studio culture

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.

Interview with:
Robin Walker (left) – Team Fortress patriarch
Erik Johnson (right) – Valve project manager

Has Valve’s dense workload resulted in a delay to improving the core Source Engine tech?

EJ: One of the things integral to Valve is that everyone’s time is their most valuable asset. Everyone treats their own time at work as a very valuable, and very limited, commodity.

I think that philosophy ties into our game engine because, if done right, it can essentially be a really useful tool that makes our job much more efficient.

Shipping The Orange Box, Left 4 Dead 1, Left 4 Dead 2 and then Portal 2 – basically in a four year time frame, has meant we have to look back at the Source engine again to improve its workflow.

We probably have under-invested in those tools in the past, to trade-off against shipping products.

How many people are working on updating the source engine and tools right now?

RW: There’s probably more people than ever. We have as many people working on our tools as we have working on a single project. So, about twenty to thirty core people. Problem is, we’re used to making decisions within the lens of a product.

EJ: A product meaning something, anything, which serves our customers. So, a product which our fans can respond to and give us feedback on. Obviously that’s not quite the same building tools internally, you don’t get as much direction in building that.

So in practical terms, what is this mass of people building towards?

RW: So I guess one part of it is improving productivity, the other is that, as Erik said, we had been working on games solidly for a long time. It left us with a lot of straightforward stuff to do.

We usually look at things in terms of products, when it comes to tools and Source, so if there’s a key technology to improve, we usually look at it in terms of what that progression can do for a certain game and customers.

When you have a product to build a technology for, you can ask yourself, well, how valuable is the addition of this technology for our customers?

But in this instance our technology had such straightforward things that needed improving that we didn’t need a product to drive directions on it.

Hardware’s moved forward, that’s one thing, and the other is the rise of user created content. That’s becoming a big deal for us. We had to address some things for that. What do we do with ten thousand content creators, who are all making things connected to gameplay, who aren’t in the same room?

Team Fortress 2 customers could be collaborating on building items for the game, to be used on a map which also happened to be built by another customer in another country as well. So we had to be aware of that reality and help those people out with enhancements of our core tech.

This is continual updates to the engine, then?

EJ: Pretty much, yeah.

Is there any ambition beyond that?

EJ: Probably not. Not that we’re talking about.

I heard a rumour that mobile was a new venture for the Source engine?

EJ: Well, for us, if you look at the profile of Valve as a company that builds entertainment software, we’ve had no mobile strategy at all. But it’s just like a bunch of other things at Valve, where we could go spend time and invest on mobile, but at least in the past we felt like the things we spent more time investing in were more valuable.

Mobile is definitely a hole in the company’s strategy – we’re not blind to that – so when we figure out a plan and think we can do something decent with it, we will.

Clearly our customers want to interact with our products through more devices than their PCs. Something like that is usually how we arrive to a decision to change things.

So you’re not ruling this out, but you’re a bit stretched.

EJ: I’m not ruling this out, I’m just being self-critical. We have no strategy here. We probably need to build one. Obviously it’s a challenge to think about how we approach mobiles, but we’re in the entertaining people business, and I think we can eventually figure out a smart way to do something.

Tell me about how both of you got a job at Valve. I hear the hiring process is gruelling.

EJ: Yes it is pretty crazy, luckily I didn’t have to go through with it. I was just one of these people that joined for a while and I never left! [Laughs] No one’s noticed yet so don’t print that! I actually came here from my employer Sierra, back in 1998, to help out with the Half Life project.

RW: I was one of the three people that made the original Team Fortress mod for Quake. So back in March ‘98 I was asked to come over and do some contract work, to basically work on Team Fortress mod for Half-Life.

EJ: You mean the Team Fortress 2 mod?

RW: Er, technically it was TF1. Or, wait, maybe it was TF2 then. Anyway, pretty quickly we were given the task to make Half-Life a good mod platform.

Why are you both still here?

EJ: The people, I’d say.

RW: Yeah definitely the people. I’ve never met a group of people like this who just want to get so much better at what they do. I find that addictive to be a part of.

EJ: There’s lots of things, for me, that are really, really important about working at Valve, but nothing is as important as the people I get to work with every day.

So if you got a job offer, for more money, somewhere else, you’d stay?

EJ: Yeah I’d be happy building spreadsheet software, as long as I could do it with the people here.

RW: Yeah. My wife would far prefer that we live in Australia instead of Seattle. Gabe’s always said that if I ever move to Australia we’d just open up a company remotely, we’d start Valve Australia, and I could work on whatever I’d want to work on.

But I don’t want to go back to Australia and work on whatever I want, I want to stay here and work on what everyone else is working on.

And I imagine being around each other makes you all better at what you do.

EJ: Yeah and the key thing is that becomes more valuable over time. Robin and I have worked together for over a decade so we communicate really usefully.

I used to be a freelance journalist working from home, doing a lot of what I wanted to do. And when I joined my company I was suddenly surrounded by people who are better than me at what I do. So, I picked things up.

EJ: You’ve just described Valve. I mean, Valve is very much a hiring company. We want the very best people work for us. One of the key tenets to doing that is always hiring people who are better at what you do than you are.

A lot of the people who work at Valve are usually the most talented people from their previous company. But it’s kind of crummy being the most talented person at the company – you’re not going to grow as much and learn as much as the people around you.

Because you have to be of a certain standard at Valve, you can’t hire en masse. You can’t load a bunch of people into a new project. This has always been why I thought there is this thing called ‘Valve Time’.

RW: I don’t think ‘Valve Time’ is a function of resources. We actually constrain our products by the resources we have. We don’t design a product and then say ‘what resources do we have for a task of this scale’.

The ‘Valve Time’ thing is a function of us just fundamentally optimising for other things over time. Which, by the way, we don’t see as a comparatively bad thing. The challenge of game design is looking at your goals, the constraints you’re under and resources you have. Within that, you have to make trade-offs. Either you loosen your goals, relax your constraints or improve your resources. At Valve, we don’t see release schedules as something we should usually make trade-offs for. We care a lot more about, fundamentally, the quality of the product or employee happiness.

When we developed Left 4 Dead 2, that was essentially an experiment in the company. We were saying, ‘hey, let’s optimise for time and see what happens’. We showed we could do it.

There’s nothing fundamentally unique about designing games that makes it particularly harder to optimise for time than it is to optimise for any other constraint. We just usually optimise for other things first.

EJ: But the company is more or less centred on optimising for invention and quality, rather than for predictability. The way we design games is not about making sure our hundredth game follows the same rules as the first. It’s about reinventing.

I am really surprised you don’t aggressively pursue the engine licensing business. Are you surprised by how few studios use your engine?

EJ: We have constraints with how much time we can spend on any given problem. We made the trade-off to build more products as opposed to spend more time on licensing.

We’ve crossed that path many times over, we heavily license the Steamworks business. But our own tech we optimise for our own products.

Don’t get me wrong, people still licence our engine, and we have filled a PS3-sized hole in that segment of our business.

That’s why I was suspicious, knowing that you’ve optimised for PS3 and you have a lot of people working on Source.

EJ: I know the Epic guys really well, and I know Mark Rein spends a whole lot of time going out there and getting people to use the great tech that those guys have. It’s just not something we’ve focused on a whole lot.

RW: I think, especially for other developers, there’s a lot of confusion about Valve from the outside. That always stems from the assumption that we work in the same way other companies work. People will imagine Gabe, or someone else senior, telling everyone else what we’re going to do and how we’re going to do it.

Really, it just doesn’t work like that. There’s a bunch of people around the company who are, individually or in groups, discussing what’s the most valuable thing they can contribute. We know how confusing that must sound.

EJ: Put it this way, we didn’t have a PS3 version of our engine because quite simply there weren’t calls from within the company from people who were interested in doing a PS3 game.

No one decided to do anything about it, and the way developers can vote to do something about it is to simply start working on it, which they’re able to do. That’s why Portal 2 is coming to the PS3, because we had about four engineers – some of which have worked here for a few years – who were annoyed that we hadn’t worked on that platform, and they said they’re willing to be the first to cross over to the PS3.

And it didn’t matter what other people thought of that. If they spent their time doing the work for it, and they made their case to their peers, then they can.

RW: Yeah if they make a case for it, even if you don’t agree with their point, if you think it’s rational and you see they’re working hard on it then you don’t get in the way of that. That’s the ideology the company is founded on.

If you understand, then, how we work – then many of the things we chose not to do become clearer. If there isn’t a group forming together, if there isn’t enough argument to go in one direction, then quite organically no one will go down that path.

So, going back to why we don’t license our engine so aggressively, if someone here really wanted to, we would.

EJ: Absolutely. If someone wanted to do the work then they’re free to. As long as it’s best for our customers.

Y’know, we essentially have no turnover here. No one really leaves the company. There’s been less than ten, ever. Most people have worked here a really long time, and most people trust the people they’re working around.

I mean, we had people here who once-upon-a-time made it clear they thought Steam was a bad idea. Obviously it wasn’t. But the people who aired those views haven’t been demoralised or punished because of that. We actually think it’s really useful to have people taking a second look at our every step.

When a group of engineers want to go off and get Portal 2 shipped for PS3, anybody who disagreed would still know we’d learn a lot from that failure. Failure is not of zero value, we don’t want to make a habit of it obviously, but failing can teach us a lot.

From what you’re saying, it seems like it wasn’t just politics that’s led to Portal 2 being released for the PS3.

EJ: Yeah, completely, and we think it’s the best console version of the game because it integrates Steamworks, and we think that’s going to be really useful for our customers.

How was it that a few people, who weren’t management, decided that a whole company was going to work on a PS3 engine and game?

RW: The information distribution process in our company is very organic. People who need information about something will go to someone else to get it. We try to automate the flow of information as much as possible, but we don’t have a weekly scheduled team meeting – people meet when they need to.

So let’s say I’m the first person thinking about porting our games to the PS3; maybe I’ve done enough thinking about it that I want to test whether it’s worth the cost. In that case I would find the person most useful to getting a sense of how much it would cost to run this project.

EJ: And you’re going to these people, like always, you want them to try and poke at your idea to make it better. We talk about cost also in terms of the time someone is not going to have if they work on this other project.

RW: Yeah and the sticking point is always about time. It’s never about the cost or the wages – time is so valuable to us that this is the most important thing to consider when discussing whether people should work on projects.

So, everyone here has this network of people they usually speak to if they want to pursue something, and you’ll trust them that they tell you straight and pick holes in your idea. That’s how you become better equipped for the thing you’re working on.

In the case of a PS3 project, I wouldn’t be speaking to management about this first. I’d talk to some engineers and see whether this is even technologically viable. Then I’d be talking to the people I’m closest to, who know me very well, to tell me if they think I can do it in a certain timeframe and quiz me on what I’ll lose out on if I pursue this.

By that point, there would be an entire group in on these discussions, and we’d have already told ourselves if it’s a good idea just by what people are saying to each other about it. If it’s clear we think it’s worth it, we start talking to people who can allow us to pursue it further.

The important thing to remember as well is that, if you’re throwing around a good idea, other people will jump onto it, quite naturally. They’ll think, oh, that’s a better use of my time. That’s how a body of people get to bring a project idea to the company.

EJ: Everyone, I should add, needs to look at how them leaving a project to jump on another is going to affect our customers. If you’re working on a really important update, and you want to jump onto the Portal 2 project, you should know it’s really important that update gets finished.

You do realise how different this system is to other studios.

EJ: We only know it’s different because we hear that it is from so many people all the time.

RW: Yeah and we haven’t really worked for any other games companies! [Laughs]

The feedback we constantly get from when new people work here is that it usually takes six months before people really understand it.

EJ: Honestly, it’s generally a lot of stress for people. People don’t get it and they think they’re not doing something right.

RW: Yeah they look for power structures. But the point is that someone external, someone above your station, can’t possibly make better decisions than you about your project. You’re the one who’s in the guts of it. People can inform you, and make you think about things, but being told how to do things just isn’t practical.

EJ: Usually it’s the urge to have someone tell them what to do. We never have that. Having someone tell someone else what to do would just be completely destructive to Valve. We’re completely allergic to that.

RW: Well, towards the end of a project, when things heat up, usually a structure emerges from within the team organically, because we need to check on every single change that’s being made to a game, and we need to kill all the bugs, and we need everything to be perfect.

But throughout the process, no one wants to make a big choice on their own. That would be terrifying. Usually, the bigger the decision, the more people are asked about it, and therefore the more it’s talked about.

Obviously it’s extraordinary to hear about how Valve operates, but I think I’ve found a flaw in this whole process.

RW: Quick! Run! [Laughs]
EJ: Tell us!

Okay, so, if you allow people to do what they want – even under the knowledge that they are remarkably talented, hard-working, self-starting team players – they will naturally gravitate to the work that excites them the most. That means, in theory, that you’ll have a studio of people doing amazing stuff, yes, but also avoiding the essential monkey work.

EJ: Actually no. We have this philosophy at Valve which permeates across the company and everyone has to adhere to: ‘What Is Best For Our Customers’. People are hired and work on the basis that they respect and understand this policy.

RW: Yeah and when we get an amazing programmer to do the monkey work, some will ask ‘well why do I have to do this? I’m super talented at this other stuff?’ It’s because that a project works out so much better when one person does everything from the ground-up, and understands everything.

And we make a habit of getting our best people to do the ‘monkey work’ as well, because usually – and this theory has been tested many times – when you get your best people to do all of a project, they usually find elegant solutions.

We have our finest artist say ‘why do I have to make world props?’ Because they’re the best and they’ll figure things out that no one else will. So our art director will design a world in such a way that no one at the company will ever have to think about making additional garbage bins ever again.

If people don’t want to do the heavy lifting, that’s their problem. We don’t want to work in an environment where anyone at the company is treated as a second-class citizen and has to do less interesting work.

Oh, well, maybe it isn’t a flaw then.

EJ: I’ll tell you what the flaw is, though. The way our company is designed, which doesn’t have a hierarchical oversight, means that if we hire someone that isn’t very good at working to our model, their ability to do a lot of bad work over time is pretty high. These people are not going to be noticed if they’re not doing a great job.

Ah, of course, and that’s why hiring people is so intense.

EJ: Exactly.

Because the impact of one person added to Valve can be quite noticeable as well. The company has set up its team structure in such away that it’s all very receptive to change.

EJ: Yeah, completely. If you run a toothpaste company where people screw caps onto bottles, you can see who did most by the end of the day. Whereas, here, if we hire someone we can only expect them to be valuable. If they’re not, it will take us a long time to notice.

RW: Because no one is watching you. But, I should add, we’d never claim that this way of working is good for any other industry. We’ve organically built this structure because we’re trying to solve our own problems. But it wouldn’t work well in companies that have to solve other problems.

EJ: I think this system is most appropriate for software development.

RW: Yeah, well, as long as the software creation has elements of creativity to it. But, certainly not for an industry like… space shuttle control.

EJ: [Aghast] Space shuttle control!? Can you imagine… ‘Sorry guys we didn’t finish that system in time’.

RW: ‘Yeah’, [laughs], ‘but we made a really innovative one… But we’ll get you your update! Just hang in there!’

EJ: [Laughs] ‘Do you have an internet connection!?’

I have a theory about Valve, and I wanted to put it past you. I think the company has entered its third phase. The first was in the late nineties, when it made isolated boxed product. The second was when Half-Life 2 and Steam came out, when it started building net-integrated games that could offer not just online deathmatch, but games that could be updated incrementally.

The third phase is an entertainment company that builds all kinds of products – comics, animations, games, merchandise – from a remarkably fluid engagement with its customers. It works solely on this rapid feedback loop, and now updates software almost on a weekly basis with a wide range of content and services. Would you say that’s accurate?

EJ: Broadly I actually agree with that. Though I would say our progression has been far more granular than three steps. Everything we’ve done since Half-Life has been about how we can interact more closely with our customers.

So we started with the first Half-Life. We put a product in a box. ‘So long, box, don’t fuck up!’ At this stage we didn’t even know who our customers were.

Steam was a massive step. We suddenly knew who our customers were and what they were doing. We found out that some of the stuff we thought we were doing right, we were actually doing awfully. We found out that some things we had not looked too closely at, people were really interested in. So by that stage, we had built a discussion with our customers.

TF2 is another step in that overall direction. It has a pretty high bandwith conversation with our customers all the time.

But, having worked here, it feels to me less stepped and more like pursuing a goal. But I see your point. And, right now, there are big shifts about what is happening with entertainment, and what customers want to experience.

Customers want to be entertained by more than just our games these days. They want to be entertained by talking to other fans, broadcasting their affinity to a product, and experiencing other forms of entertainment associated with that franchise. They want to be able to watch a movie, or read a comic, and they want a reward structure in place for that as well. And they want to create their own content for it. You have that all tied into a system that essentially makes it an all-encompassing entertainment experience.

I think we were focusing on this for a long time, it’s just that we hadn’t been as good, or well-equipped in breaking down the problem.

RW: Yeah I agree with your point. I should say that when Steam went live, the most exciting thing was the potential to engage better with our customers. I mean, holy shit, we can send out four updates to Team Fortress 2 in two days. The data we get back from Steam is just incredible.

EJ: Yeah when I first saw Steam at Valve, I honestly thought it was a way we could compete with retail distribution. No matter how many times you told people that was an interesting problem to attack, they didn’t believe you.

When we launched Half-Life 2, though let’s not talk about the first two days [sighs], the thing I had the most value out of it was knowing that we could deliver all of this content to our customers, right now.

And the risk level in shipping a product went down so much, it really has had a positive affect on game development.

RW: In fact we have been able to modify a game before its release based on how people respond to marketing.

When customers watch our trailers and say ‘that feature looks awesome, I can’t wait to use it to do X, Y and Z’, sometimes we haven’t thought about those ideas at all, so we’ll update a game pre-release to add stuff in that people are excited about. That’s the power of Steam.

EJ: Taking an entertaining view on something people have about a trailer and turning it into a reality for people is petty amazing.

TF2 is a big deal today, but I think that when people look back on the history of Valve it will be seen as an even bigger turning-point. Primarily because now it has introduced the microtransaction model, which Zynga has cracked for one market but is still up for grabs in the hardcore space.

EJ: I have a tough time talking about the whole in-app payments and microtransaction system, specifically because there’s such a customer-hostile approach to it.

The theory a lot with DLC and microtransactions and the rest of it is ‘we are not extracting enough dollars from our customers, so let’s find another way of extracting an optimal amount from them’.

In that regard you’ve kind of already gone off the tracks a little bit. The ideology behind microtransactions should always be ‘how can I entertain my customers as much as I can’.

RW: I don’t know how to react when someone says Team Fortress is going to be seen as a really important milestone in Valve’s history. TF2 is where it is today because every day people come into work and think about how they can improve the game’s experience for its customers.

I take it that it will be a long time before TF3 is even considered, and I’m only saying because TF2 will be supported until its last customers are still there.

RW: The choice is that work on TF3 would solve some problems but create other ones. Maybe the shift to that new project could be in a few years. Maybe it could be tomorrow. Maybe it was six months ago and we should have made the decision to shift to a new project, and we missed it.

TF2 has changed significantly since it was first shipped. Customers who consumed its marketing message three years ago might not have been interested in playing but might be really interested today, because the game has changed so much.

So maybe there is a value in having a fully updated version called TF3, to again communicate to those fans that the series has moved on. Maybe, I don’t know. Doing so would create problems of its own.

Are you guys excited about Black Mesa Source?

RW: Oh god yes! We’re all super excited. We would be super exited even if it didn’t look so amazing.

It’s really, really, really hard to create stuff, to make any game at all. So when we see people out there who’ve managed to make something, and they release it, it’s just staggering to us.

So when the community get together and build something that looks so cool, I mean, who wouldn’t be excited. The whole Black Mesa thing is staggeringly interesting. I can’t wait.

Have you guys spoken to them much?

EJ: I’ve traded a few emails in the past.

RW: Yeah same here. One of the tricky things is with those mod groups is that they’re big and amorphous and people come and go and so it’s a bit harder to maintain a single contact. But a group of people together making something they really care about sounds a lot like how we do things here.

I was thinking about describing Valve as a giant professional mod community for my article, actually.

RW: The whole idea about constantly updating our games originally came from a mod community ethos. The best mods in the early days, like Counter-Strike, Team Fortress, Day of Defeat and Garry’s Mod – these were all things that were constantly updated.

And when you realise these were the most popular mods and see that they update all the time, you think, yeah, maybe there’s something remarkably valuable in building your product iteratively in front of your own customers.

I was really worried actually that, because of this direction you’re going down, you will no longer be working on big five-year projects.

RW: I think the thing is, Half-Life 2 didn’t really take us five years. It took us two and a half years, and another two and a half not really knowing what the hell we were doing.

EJ: We were dumb then.

I find your stonewall ‘no-comment’ responses to the next Half-Life project interesting, because with every other aspect at Valve, you maintain a discussion with fans right until the very end. But before you announce a project, there’s stone silence. Is that to get people excited?

EJ: Oh no way. We’re not trying to spin people to get them super excited. We’re fans of games too and we know what it’s like to be in their seat. Part of it is that we try and be very nimble and fluid with how we make games, and sometimes that can be at odds with how we talk to our customers.

RW: Yeah we’ve been in situations before when we’ve said, hey, we’re going to make X. And then we don’t make X and people ask us why. And we say, yeah, here are the list of reasons why we couldn’t make X, and they say, screw that, you said you were making X!

EJ: And that’s really damaging to our customers. We try to be very careful about setting expectations about what we’re working on. I’m always going back and forth about how transparent we should be, because the reality is that we’re changing our mind all the time about what’s best for our customers.

But the problem is these changes aren’t always communicated to the customer, they just see snippets of info here and there, so you want to be clear what we’re talking about. It’s a really hard problem.

The silence is not about getting people excited about something, it’s about protecting them from an assumption we have about a project that could, in the end, go another way.