Promoting Valve’s quirky employee handbook as a manifesto for unfettered workplace creativity set Gabe Newell on a collision course when he hired the legendary geek goddess

I’m not going to say very much about the interview in the video below (which, by the way is by far the best ‘exit interview’ I have ever seen) or Jeri Ellsworth, or Valve: instead, I want to give my take on a problem at Valve that she talks about, then I want to consider how it might reflect upon the tricky relationship between management and innovation and then I’d like to hint at a potential solution.

But is Valve management all about flatness, or something else?

I’ve been sticking with flatness as a name for what Valve does differently about management, but the term really applies more to the structure of an organisation, rather than exactly how it is managed.

A single owner with maybe a handful of managers working at one level down from the top, running hundreds of staff, is pretty flat, but just that description alone doesn’t tell you anything about how ‘traditionally autocratic’ or alternatively how ‘consensually Valve-like’ the management style happens to be.


What characterises Valve is that it purports to be ‘hyperflat’: not just by having ‘fewer layers of management’ but instead having ‘no layers of management’.

That doesn’t mean that there is nobody in charge (somebody, after all, had to fire Jeri, and from what she says, it sounds like it was Gabe Newell himself).

But it does mean that for most decisions, the staff have to make those decisions among themselves, by coming to a consensus.

The reason why this is a rare arrangement may seem blindingly obvious, but I’ll spell it out anyway: it’s ‘commercially unattractive’ because it involves more people in decisions, involves more time spent deciding things, and thus costs more to get things done than ‘bosses just giving out instructions’.

It just seems wasteful, slow, impractical and unnecessary.

Management by committee seems like something you do only if you are formally obliged to do it (e.g., in public bodies).

So why do things in this seemingly stupidly cumbersome way if you don’t have to?

It’s the innovation, stupid

If you are studiously trying to avoid the innovator’s dilemma, which tells us that large organisations eventually become ‘too big to innovate’ and that this innovator’s block comes from things like:

‘Too many layers of management’

Even when you don’t need to form a committee to get something done, you still have to get so many people in so many layers ‘on board’ with a new idea : the layers introduce a potentially enormous time delay, even if the new idea is extremely welcome and seems to pose no risk to existing arrangements

‘Risk aversion’

The larger the number of people involved in the existing operation, the more there is at stake if a new option is pursued.

This is also know as ‘business model preservation’ where a new business model (or even a tweak to an old one) has the potential to undermine an older one.

The kind of courage that is required to risk undermining an old business model with a new one, especially if the new one is essentially innovative and untried, is quite rare in the management of large organisations, which are typically not very flat.

Gabe Newell opines that the creative nature of games development opens up the option for him to hire the very best and most expensive staff in order to ‘max-out’ the innovation.

As a result, Gabe is ultimately hiring those whose self-management is already likely to have a good track record, but whose unsuitability to being managed by others is likely to be a recipe for trouble (and a potential barrier to recruiting the very best).

Just let them manage themselves and each other

The logical answer to the problem that this poses is essentially a ‘hands-off’ approach to management.

The problem that I’m exploring in this article, is that despite attempting to prevent the problems of management by eliminating the deployment of managers, it turns out that unless you take specific measures to address the problems associated with ‘peer management’, you will find that ‘tacit management structures’ are likely to emerge and these are just as likely to have a detrimental impact upon innovation as conventional top-down structures can have.

An irresistible force

For a moment, I’ll just focus on one counter-intuitive aspect of innovation that pops up in the Ellsworth firing story.

It’s an incident which combines the implications of removing restrictions upon creative freedom with the impact this has upon the options available to an innovator and its repercussions for peer-management.

Among all the other things that they need to be able to do in order to innovate, an innovator needs the freedom to do the one thing that innovators are not supposed to need to do.

In addition to working together collaboratively with fellow innovators, they also need to be able to freely exercise the option to embrace something more akin to ‘non-innovativeness’ on the recruitment side of things.

You can do this anywhere else, but not at Valve

They need to be allowed to engage and deploy human resources who do have clearly pre-determined functions and to engage them to undertake tightly pre-defined tasks (in other words, they need to be allowed to override their ‘choose your own roles and goals’ imperative) in order to get certain things done in a reasonable time-frame.

Valve would not let Jeri do this (she wanted to employ a professional machinist on her project and her peers fought back) and their response was symptomatic of a bigger problem, which relates to the kinds of ’emergent constraints’ that paradoxically result from any well-intentioned attempt to systematically remove all ‘external’ managerial constraints from a creative enterprise.

An immovable object

There are very few places in the world of work where an implacable insistence upon the kinds of freedom from ‘hierarchical instruction’ is an incontrovertible article of faith, published (as The Valve Employee Handbook) and paraded as a defiant example of ‘what makes us different’ to all those unconvinced.

Creativity needs constraints

Taking constraints (in terms of pre-assigned or subsequently imposed goals and roles) out of the loop when you are working with others (especially when you haven’t worked with them before) inside a larger ‘open plan’ group can easily go seriously wrong.

I’m going to take it on trust that at Valve there is plenty of successful peer-managed creativity (in fact I’ll need to disclose here that I am quite convinced that Valve is actually one of the most important sources of innovation in existence, and that I don’t play videogames) but from what Jeri says, it might be worth spelling out what happens when things turn sour in peer-managed environments.

‘Creativity theatre’

When it does go wrong at an institutional level, ‘peer-managed creativity’ can be ‘enacted’ as a kind of (narcissistically?) ritualised activity (where any emerging irreconcilable creative differences between participants result in endless rounds of confrontational and/or passive-aggressive ‘non-negotiative posturing’).

I’m not saying this is what happens at Valve, but if things did go wrong in the way that Jeri describes, then something inherent in the difficulties of combining peer management and the creative process may well be worth putting under the microscope at at Valve.

Constraints come in through the back door anyway

The constraint that is introduced when there are no bosses imposing constraints on what you do or how you do it? ‘the tyranny of peer consensus’.

You can be told that you can do whatever you want, however you want to do it, and with whichever members of staff you want to work with, but even if it is an entirely solo project, at least some others in the organisation have to share and support your aims in order for the project to go forward.

If this doesn’t happen when support is unavoidably needed, then all the top-down-granted individual creative freedom in the world will count for nothing and you will be just as creatively hamstrung and frustrated as you would be if you had a boss who just expected you to follow orders.

Peers can turn into bosses…

The harder you push back against any strategic lack of peer support for your project, the more peers will see you as attempting to undermine their creative freedom as embodied by exercising their option to reject your ideas.

…or enemies

If you successfully garner support from any sympathisers against such rejection, you may ultimately be seen by those who rejected your ideas as being the cause of an inevitable impending upheaval.

And so we see the beginnings of something supposedly impossible: a seemingly free choice of what to do and how to do it transforming itself into a comparatively rigid and impositional hierarchy, hell-bent upon nipping ‘the wrong kind of innovation’ (or innovator) in the bud.

Moderation (or at least moderators) in all things

As Jeri discovered, for the kinds of freedom that she needed, Valve’s ‘flat’ system could still sometimes work out to be just as unsuitable (for the things she wanted to work on, in the way she wanted to work on them) as a traditional management process would have been.

This kind of menagerie needs supervision: not by old-fashioned paternalistic taskmasters, but by a type of creature whose job has nothing to do with deciding what is being done, or how, or when.

The kind of special resource that is needed in order to maintain creativity and innovation in a peer-managed environment seems closer to that provided by a really good moderator in an online forum.

Moderators are not acting as line managers or executives: they are not responsible for recruiting and tasking participants with operational requirements, or setting strategic priorities or targets, or measuring performance in terms of participants’ attendance, punctuality or productivity.

They are not responsible for commissioning or creating ‘content’, and so the precise nature and overall quality of the content that is produced by the participants is outside of their direct control.

Nonetheless, they do retain an important responsibility for key aspects of how participants interact with one another, in a way which is ultimately indirectly reflected in the quality, innovativeness and quantity of the content and the overall sustainability of the collective endeavour and the cohesiveness of the community that uses the resource.

Jeri says she will be taking the flatness of Valve into her new venture, but from my perspective, those that want their operations to be both flat and large (i.e., the typical high growth innovative startup project) need to be be warned:

Trouble in flatland

Flat organisations may facilitate innovation, but they need to do special things to maintain both the flatness and the innovation over the longer term.

Things need to be done which have far less to do with merely insisting upon flatness as a route to innovation than they have to do with systematically attending to all too easily overlooked nuances of human relations in an intensely creative environment.

Creative types have a well-deserved reputation for tremendous innovation but poor manageability: getting them to manage each other (in a way that feels to them as if they don’t have a boss, but they still get things done) sounds like an ideal solution to this seemingly intractable problem.

It sounds as if introducing the very freedom which characterises their best work (and the spirit which inspires it) will prove to be the basis for putting together an idyllic collaborative workplace for creative talent.

In practice, however, the very thing which would have made them unable to sustain their creativity if they were managed under ‘authority’ can and almost certainly will insidiously manifest itself when they work together as peers.

Cliques will form (‘this bunch just tend to like working on the same kind of project’) consistency will naturally emerge as a subliminal, defensive, ‘organic’ response to anxiety concerning unwelcome interference by outsiders, and innovation will ultimately be constrained within the ‘acceptable’ parameters set by such ‘informal network subcultures’ (insiders will inevitably just call those parameters ‘good taste’).

The innovator’s dilemma can still flourish in flatland, because peer-management is a social phenomenon which introduces its own innovation-impacting dynamics.

The social cohesion inherent in peer-managed groups inevitably introduces boundaries which are just as likely to undermine the long-term sustainability of innovation (by eventually developing consistency resulting in stagnation) as they are to propel it at its inception and reinforce it in the short term.

Peer or hierarchy, which is best for startups?

Small startups can be peer-managed or they can be hierarchical: even among the startup’s founders they may be a flat, level playing-field, or instead, one or more founders may defer to ‘the leader’. John Lennon in the Beatles? A really creative group, seemingly not peer-managed.

Gabe Newell has hundreds of ‘artists’ working on scores of projects at Valve, but his aim is to use peer-management as a way to ensure that each project retains the innovativeness of a startup.

By the way, leadership and peer-management are not mutually exclusive: a leader can lead by persuasion and steering consensus, rather than by securing obedience to instructions.

My suggestion is that those boundaries and the staff which they enclose may need to be systematically interacted with by specially tasked operatives who are neither ‘working on projects’ nor ‘responsible for project progress’ but who are on the lookout for barriers to socially sustainable innovation, at least one of which seems to have been crashed into by Jeri Ellsworth.

The video below is the first segment of a six-part set and I’ve set it start playing at the point where the discussion turns to her recent experience at Valve.

Jeri goes on to describe an augmented reality gaming system called CastAR developed by her new company Technical Illusions which happens to be at the center of the interest surrounding her departure.

It is worth me putting a marker down here by saying that despite its current somewhat ramshackle appearance, don’t be fooled: CastAR looks to me to have the most potential of any new user interface technology other than the Oculus Rift (and I’m not even sure whether that’s an understatement).

You will find that there are some glitches in the videoconferencing software (but they only last a few seconds) and the sound and picture quality of the video are definitely not the best, but it doesn’t make the video any the less worth watching.

The second video here is probably the most widely known, where Jeri takes a vintage Commodore 64 PC and shows you how she engineered it into a combination of bass guitar and synthesizer.