Cultivating a culture of experimentation
How to truly foster creativity, originality and innovation in game development
The playbook for “innovation culture” typically looks like this: you set aside 20% “creative” or “pet project” time, run quarterly hackathons, put up a poster about psychological safety and maybe schedule a TED talk about “failing often and fast” at the all-hands. But two years later, the game the studio shipped was somehow “just okay”, because originality was not really, truly fostered in the team. It was more like decoration in the shape of posters and programs that, technically, check all the right boxes.
The studios that ship original work do something different than that. What produces a Hades, or an Elden Ring, or a Cuphead is not so much a program, but a stack of interlocking practices, rolled out slowly and protected frantically.
This article, which leans heavily on scientific research and empirical evidence, is about the creativity-fostering conditions that separates studios that consistently produce original games from the ones that talk about it. Six conditions. They are mutually reinforcing, which means implementing one of them in isolation doesn’t really do all that much. They map closely to what the research actually says about how creativity works in teams.
The decoration problem
In an organization whose default behavior is not to support creativity, “Innovation Fridays” just don’t really move the needle. A team that gets one day a week to “be creative” (as if you could summon creativity on command) spends the other four days being told to hit a milestone, and the creative day produces prototypes that no one has the budget or executive sponsorship to actually build. The intention is good, but the work is done in a neat little sandbox the rest of the company just won’t really have a chance to get behind.
I’m not speculating about this. David Burkus’ The Myths of Creativity walks through the empirical record on brainstorming, idea contests, suggestion boxes and the rest of the standard interventions, and the data on most of them is frankly not encouraging. These practices do reliably produce more ideas. What they do not reliably produce is more shipped, realized originality. Teresa Amabile’s componential theory of creativity, refined over a whopping thirty years of studies, highlights the underlying reason: creativity is driven by intrinsic motivation, and most “innovation programs” are extrinsic motivators. It’s that simple. A hackathon with a prize is a contest. A 20% time policy with a manager checking in on it is, in a way, a deliverable.
What actually works is changing the organization itself.
1. Independence is a precondition
Swen Vincke at Larian has said, more than once, that the studio’s independence is what allowed it to take wild chances on Baldur’s Gate 3. He’s also said he was willing to sacrifice anything, including rewriting whole engines (which I can seldom imagine being a good idea), to ship the game in the state he wanted. Those two statements are actually connected. He could rewrite the engine because no publisher was holding him to a milestone payment schedule that required the old engine to be used or that didn’t account for that technical body of work.
This may sound a little ideological but it’s not. Studio independence is a structural precondition for being able to kill a prototype, change a major system or extend a production cycle without owing someone an explanation. The moment you’re on the hook for milestone with the publisher, every “we should try a different approach” conversation gets routed through a contract amendment. After the third or fourth amendment, the team sadly learns to stop suggesting different approaches. That’s not conducive of innovation.
You can see something similar at FromSoftware, which famously released the first couple of Souls games to somewhere between mediocre and okay commercial reception, and was protected from publisher pressure long enough to build the franchise that now defines the studio and that spawned an entire new genre. You see it at Klei, which stayed self-funded until 2021 and used that period to make Mark of the Ninja, Don’t Starve, Oxygen Not Included, all exceptional and very different games from each other. You see it at Supergiant, which has remained independent across four different IPs and has explicitly refused to run parallel projects so that one game gets the team’s full attention.
You can see the same pattern in the research. The seminal Self-Determination Theory, developed by Edward Deci and Richard Ryan over four decades, identifies autonomy as one of three psychological needs that drive intrinsic motivation.
Without autonomy you can keep people busy, but you can’t keep them creative. In a game studio autonomy is independence.
But alas, independence comes at a price. If you walk away from term sheets you genuinely need, sometimes a competitor with publisher backing reaches a market window before you do, or sometimes you just run out of money and have to fold. What you get in return though is the ability to kill a prototype that isn’t working without owing anyone an explanation, and that ability is the foundation everything else in this article depends on.
2. Clearly led small teams
Klei had eleven people in 2009, thirty-five by 2013. Supergiant has stayed small enough across four projects that “no parallel projects” is a stated policy. House House made Untitled Goose Game with four people. The studios that ship original work tend to be smaller than the work hints at, and they tend to defend that size against the natural and/or financial pressure to grow.
It comes down to coordination overhead, which rises faster than headcount. The marginal hire on a team of forty is quite possibly making the team less responsive, not more. Small teams can hold the game in their heads. Large teams have to externalize the game into documents, and in many cases, the documents become the game, which is not great.
The thing that goes with smallness, and that is equally important, is clear creative direction. Hidetaka Miyazaki at FromSoftware has said the priority on Elden Ring was “a feeling of exploration above everything else.” Jenonva Chen at thatgamecompany has said his teams “start by mapping out what the game should make the player feel,” and that mechanics derive from that target. These directors are not running design-by-committee. What they’re doing is running small teams with a strong creative spine, which is a very distinct and clearly effective structure.
Pixar is perhaps the clearest benchmark. Ed Catmull’s Braintrust, described in Creativity, Inc., is the standard version of how this combination works in a creative production. Very experienced peers review the work in progress, give brutal candor about what isn’t working, but have no authority to mandate changes. The director keeps the keys. The team stays small enough that the director can actually direct. The candor stays high because the director isn’t being threatened by it. At least in principle.
Valve is the test case for what happens when you skip the “clearly led” part of this equation. The flat-structure handbook is one of the most-circulated org documents of the last fifteen years. The reality though, as former hardware lead Jeri Ellsworth and others have described it publicly, is that Valve’s flatness inadvertently produced a hidden layer of powerful management plus chronic resource starvation of risky projects on the game development side of the company. Flatness without strong creative direction can easily degrade into politics, where the most well-connected people, and those with the loudest voices, pull resources toward sure-thing work, keeping the genuinely risky work from ever reaching the surface.
The studios that ship original work tend to hold these two lines together. Kept small enough to actually iterate and kept centered on someone who can actually provide effective, creative direction.
3. Ritualized visible failure
In 2007, Tim Schafer at Double Fine introduced what he called Amnesia Fortnight. Every employee drops their current work for two weeks and joins a small group prototyping a brand-new game. Several of those prototypes became shipped Double Fine titles. The ritual has been run repeatedly, sometimes publicly with fan voting. Schafer credits the inspiration to film director Wong Kar-Wai, who took the Ashes of Time crew off to Hong Kong mid-shoot to make Chungking Express and Fallen Angels for fun. Both of those films received better critical reviews than Ashes of Time.
Amnesia Fortnight isn’t really about the prototypes. What it does is institutionalize prototyping as a public, visible, scheduled activity that the whole team participates in. Everyone knows when it’s happening. Everyone knows the work will be shown. Everyone knows most of what gets made will not become a game, and that’s fine, because the cultural understanding is that this is how the studio makes games.
Supergiant does this through Early Access. Hades shipped in Early Access for two years, Hades II for about eighteen months. Greg Kasavin has said the team’s working principle was “not being too precious about these things when they exist in service of a cohesive whole.” The team killed and replaced systems in public, with a paying community watching. The visibility was critical for them and their process.
The most extreme form of this that I know is Supercell. They are a mobile company so they’re outside of my usual game development analyses, but I find their ritual to be pretty darn cool. When Supercell kills a project, the team that made it pops champagne. The way they frame is that it’s not failure itself that they’re celebrating, but the courage to take a swing, and to learn from it. Their CEO Ilkka Paananen said in 2023 that the studio had killed thirty-plus games against five hits over more than a decade, including Everdale, which had a real soft launch and a real player base before the team called it. The champagne ritual is the cultural mechanism that transforms the kill from a failure event into a normal step in the creative and product development process.
The Edmondson research on psychological safety, particularly her work on what she calls intelligent failure, describes what these studios are doing in academic terms. The Frazier meta-analysis confirms that psychologically safe teams produce more creativity, more information sharing, more learning behavior. Keith Sawyer’s research on group flow, summarized in Group Genius , adds that jazz ensembles almost never reach group flow during rehearsal, only during performance, because flow requires the real possibility of public failure.
A team that can’t fail visibly is less likely to reach the conditions that produce the good stuff.
A decent barometer for culture is whether a killed prototype is treated as data or talked about as wasted budget. If it’s the second, no number of Amnesia Fortnight clones will fix it.
4. Constraints
Russian composer Igor Stravinsky said you cannot create against a yielding medium. Art needs resistance. Patricia Stokes’s Creativity from Constraints builds a case from the lives of artists, designers and architects: the breakthroughs tend to come from self-imposed restrictions, not from blank canvases. Cromwell’s 2024 review in Organizational Psychology Review pulls together the empirical work and finds an inverted-U shape.
Too few constraints and the team flounders. Too many and the team can only execute. The middle band, the goldilocks zone where constraints force invention but leave room to invent, is where creativity peaks.
Studios that produce original work brag about their constraints rather than apologize for them. Studio MDHR built Cuphead under the explicit constraint of hand-drawn 1930s animation. Every visual problem became “how do we solve this in cel-style.” Think about it: most of what makes Cuphead recognizable is a consequence of that single constraint. House House made Untitled Goose Game with four people and a small scope, and the small scope is part of the appeal. Jenova Chen describes thatgamecompany’s design discipline as a Japanese garden rule: the design is right when nothing more can be removed.
Another constraint that may fly under the radar, but that many great studios use, is “no parallel projects.” I touched on this briefly already, but Supergiant has remarkably held this line for sixteen years. And it costs them. They can’t scale, they can’t run a big experimental side project, they can’t hedge. What they get though is a team where every person is working on the same thing at the same time, where decisions about that thing don’t have to negotiate with another team’s roadmap and the game gets the studio’s full creative attention. And that shows in spades in the results.
Studios that don’t hold this “no-parallelization” constraint sometimes end up having a sequel, a port, an expansion and a new IP in production all at once. The team gets split four ways, the new IP gets whatever attention is left over after the safer projects are staffed, and nobody pushes for any serious creative risk on it in review meetings because everyone is already stretched to the max.
5. Time architected so pressure and exploration alternate
Amabile and her team analyzed over 9,000 daily diary entries from people doing creative work and published the results in HBR as “Creativity Under the Gun” (also here) . The finding most relevant in this work is the four-state grid below for how creative work actually behaves under different time conditions:
High pressure with focus is “on a mission,” and creativity holds up
High pressure without focus is “on a treadmill,” and creativity collapses
Low pressure with curiosity is “on an expedition,” and creativity peaks
Low pressure without curiosity is autopilot, and nothing useful happens
Something that should make every producer take a moment and assess is the data on perception. Workers under treadmill conditions report feeling more creative on those days. But they’re actually producing less. The subjective experience flips the objective output, which is one reason the standard AAA crunch model produces so few original ideas. It generates the feeling of intensity without the actual creative work.
The studios that ship original work tend to deliberately architect time as alternating between mission states and expedition states. In principle, Larian rewriting an engine rather than ship a broken Baldur’s Gate 3 is a mission state. High pressure, sustained focus, the team knows exactly what they are trying to do and why. Supergiant’s Early Access is closer to an expedition state. The pressure is real, but the curiosity is the point, and the team is allowed to follow what’s interesting in the playtests rather than crunch toward a fixed feature list.
What you seldom see in studios that produce original work is the full-year treadmill, where the milestone keeps shifting, the priorities keep getting reshuffled and the team is continuously stuck in the high-pressure-without-focus quadrant. The studios I worked with that lived in that quadrant produced exactly zero original mechanics in the years they spent there. They shipped, and what they shipped was competent and forgettable.
The type of slack time that actually works isn’t “Innovation Friday”, but the protected expedition period, which is a stretch of weeks or months where the team doesn’t have a milestone, doesn’t have a deliverable, doesn’t have a manager checking in on what they made. They read, they play, they prototype things that have no roadmap claim. They invite curiosity and creativity to flourish. Studios that say they have this often don’t, because there’s always one more milestone. I wouldn’t be surprised if the studios that do can point to specific shipped games that came out of an expedition period.
6. Kindness in hiring with actual behavioral consequences
Tim Schafer has said, on the record, that he doesn’t hire mean people. He hires for kindness alongside talent and treats kindness as a precondition rather than a tiebreaker. This is a pretty important component of the stack that I would wager a lot of studios don’t think enough about.
Psychological safety doesn’t last without norms that protect it, which is what Edmondson’s 1999 paper and the Frazier meta-analysis both find. A team where bad behavior is tolerated as long as the results are there isn’t a psychologically safe team for everyone. It’s psychologically safe for the person behaving badly, and increasingly unsafe for everyone else. Over time, the people who could be honest about a struggling prototype stop being honest, because the loudest person in the room punishes anyone who slows them down. The candor evaporates and the Braintrust becomes theater.
I worked at a studio a while back where this was the modus operandi. Bad behavior was totally acceptable as long as the results were there. The senior people who delivered were given enormous behavioral latitude, and the people who flagged the cost of that latitude were treated as a problem. The official line, as you would expect, was that we had a “culture of excellence”. But the actual culture was defensiveness, internal politics and an inordinate exodus of senior-level talent who had options elsewhere. The games that came out of that period were technically impressive and creatively unremarkable, and the people who could have made them more interesting had either left or learned to stop trying.
This is also where the defining Project Aristotle story (which I have referenced and advocated for before) needs a little qualification. Google’s claim that psychological safety is the most important factor in team performance is famous, but, as Dennis Adsit and others have pointed out, the study has never been independently replicated, the methodology was never published, and the social-psychology field has a replication crisis that should make any single unreplicated finding suspect. What is solid though, across Edmondson, Frazier and the broader literature, is that safety reliably correlates with creative output across many studies, and that safety requires norms that exclude the people who would otherwise destroy it.
The studios that ship original work are not soft. They are kind, which is a very important distinction. They fire for behavior, not just for performance. They protect the people who would otherwise be eaten by the loudest person in the room.
The cost is that you sometimes lose a productive talent who can’t operate inside those norms. And the benefit is that the rest of the team can.
The conditions are synergistic
The six conditions don’t act independently, which might be easy to miss when reading a list like this. They’re a system. Take any one out and the rest might just start failing. Independence without small teams might give you a director who’s drowning in scope. Make failure visible in a team without intrinsic motivation and you get Braintrust theater. Pile constraints onto a treadmill schedule and the team burns out before the constraint forces invention. Tolerate bad behavior in an otherwise kind hiring policy and the safety collapses for everyone except the loudest person.
My observation from the prep work for this article is that studios that shipped original work implemented the stack as a stack. I would guess that most of them didn’t do it entirely consciously. They probably figured out one piece, then the next piece, then the next, over many years and several near-death experiences, and what they ended up with was a whole organization tuned to keep the conditions in place. The rituals everyone copies (Amnesia Fortnight, Early Access, the Braintrust, the champagne) are just the visible elements. The invisible and truly important one is the structural commitment underneath, which is what makes the rituals work.
If you are building or running a studio and you want cultivate a culture of experimentation, more than wonder which program to implement, you should think about which condition you can credibly nurture first, which one you’ll need to foster next and which one you are willing to pay the cost of maintaining when the cash gets tight or the schedule slips or the loudest person in the room starts arguing for an exception.
If you want to go deeper
A short list:
Ed Catmull, Creativity, Inc. (2014). The operational manual. The Braintrust chapters are the most relevant.
Amy Edmondson, The Fearless Organization (2018). The clearest book-length treatment of what psychological safety actually is, and isn’t.
Keith Sawyer, Group Genius (2007). The empirical work on group flow, with a lovely working knowledge of jazz that informs the whole argument.
David Burkus, The Myths of Creativity (2013). Brainstorming and innovation contests, with the actual data on how poorly they perform.
Patricia Stokes, Creativity from Constraints (2005). The case from artists’ lives. Stravinsky’s tonality, Picasso’s blue period, Frank Lloyd Wright’s site rules.
Five books gets you most of what the literature has to offer. One way or another, the studios are doing what the books describe.
What’s your take on how to best foster creativity? What have you seen work? Leave a comment.
If this was useful, you might also like:
I'm Sebastian. Twenty years across AAA studios (Riot, Crytek, EA), indie and founding roles. I write at Game Studio Unlocked because I want to help the next generation of indie game studios do great. I also work directly with a few studios at a time as a fractional COO and executive producer. More at gamestudiounlocked.com or on LinkedIn.