Why Our Game Flopped (Part 2)
Analyzing a failed launch
This is Part 2 of my series on why our game flopped. Here for the first time? Go back and read Part 1 where I discuss the power of a strong fantasy.
Hello to all of my new subscribers. And thanks to The GameDiscoverCo newsletter for their thoughtful notes on my first post.
Quick note: If you’re making games as a hobby, for a game jam, as a student project, as an art project, or as a passion project, then this post is not for you. You should ignore it and set out to make whatever you want to make.
This post is for people interested in using Steam data to help them better understand the market, and in turn, set out to make great games that are more likely to succeed.
The most important decision you’ll make in developing games comes before you write a single line of code or sketch your first piece of concept art.
Genre.
Pick the wrong genre and your game is dead on arrival. The bad news? You probably won’t realize it until your game is released. Only then will you finally turn around, unblock your nostrils, and see the wretched, rotting corpse you’ve been dragging along this whole time.
But hey, at least it’s your rotting corpse. Might as well share it with the world.
Going genre first feels sacrilegious for a lot of developers. They’ll say: game development is art, and art is about the freedom of creative expression, the totally raw transmission of my unique ideas and feelings.
That’s your ego talking. Probably best to kill it if you want to make games that sell.
I made this mistake not once, not twice, but three times in three years.
Reggie James talks about this in his post Distributed Determined Design. “Channel” in the quote below is referring to “distribution channels.”
I’ve come to believe it is absurdly hard to change channel behavior. In fact, you simply won’t. You either understand how a channel operates and win, you understand how a channel operates and your ego gets in the way and you lose (many such cases), or you don’t understand how a channel operates (also many such cases).
I believe gaming is generally ahead of the rest of software in more ways than one, and when it comes to understanding mobile distribution - it’s no different. User acquisition for mobile games has one channel and one model → paid ads.
Paid ads in mobile not only serve as user acquisition, but user research in what to build. “Distribution Determined Design”.
Our distribution channel was Steam. And fully understanding Steam and the surrounding ecosystem would be key to our success.
Genre rules on Steam
Genre really matters in games. More so than in movies, music, or other media. Games ask for much more of your time and attention, especially PC games. And players on Steam want to know exactly what they can expect before committing their time and money.
Every game on Steam has a set of tags. These sit right at the top of its store page, just below the description. Tags are set by the developer and can be tweaked by players. They range from broad genre categories like Adventure or Strategy, to more specific ones like Dating Sims or Automation, to even more specific thematic descriptors like Zombies or Trains.
Tags are combined with capsule art (Steam’s equivalent of a movie poster or album cover), the game description, screenshots, and a trailer to give players a clear picture of what they’re in for.
Many players skimming your Steam page want one simple question answered: Is this like that other game I love? Players tend to think and shop in terms of genre.
But let me be clear, this is really a vibes thing. It’s part of the dark art of a good Steam page. Players aren’t sitting there analyzing every detail of your page. They’ll feel it out. And if something feels off, they leave, the game is not for them. They’ll just know.
Every year, Chris Zukowski from HTMAG breaks down sales data from Steam and puts together a summary of the top performing genres. The top five genres from 2024, based on the percentage of games in that genre that reached over 1,000 reviews, were: Open World Survival Craft, Farming, Roguelike Deckbuilder, Management, and Simulation.
And these aren’t just fads. Some genres come and go but the top performing genres on Steam basically don’t change year after year.
PC gamers want games they can lose themselves in for hundreds, if not thousands, of hours. Deep systems, long-term progression, something that hooks them and keeps them coming back.
A good chunk of the best-selling indie games on Steam don’t even have you controlling a character. No intricate movement systems, no cinematic cutscenes. Just a cursor clicking on menus, managing systems, and watching numbers go up.
Three years ago we set out to make our first game. What did we do? We picked three of the worst performing genres on Steam.
I wish I were joking.
Let’s dive in.
Mistake 1: A VR Walking Simulator
Our first attempt was in VR. It made sense at the time. We had been working in VR for years. Specifically, we set out to make a walking simulator in VR. This is a genre with little to no gameplay mechanics, just environmental storytelling, strong atmosphere, narrative over gameplay. It was niche, but there were hits. This felt like a good entry point into for us. Little to no mechanics.
We looked at games like Gone Home, What Remains of Edith Finch, Firewatch, and Tacoma. These pulled in anywhere from $1.2M to $15M gross on Steam (we failed to look at over 3000 of the other games tagged with walking simulator on Steam that hadn’t succeeded.)
I was personally drawn to the slow, meditative exploration in Death Stranding and wondered if we could capture that feeling on a smaller scale.
Our previous VR projects were built by 3D scanning physical sculptures and artworks and reconfiguring them in Unreal Engine. We were excited to use the same technique in games, so we brought in a visual artist (sculptor/illustrator) to help with worldbuilding.
At night, I dreamt of the worlds we were building. Eight islands, each with its own strange history, connected by a massive ship that the player would use to travel between them.
A few months in to making our prototype, reality hit.
We didn’t have an efficient enough pipeline to turn a massive amount of 3D-scanned sculptures or other artworks into game-ready assets. We struggled to find freelancers. We didn’t have the funding to commit to full-time hires.
And then there was the market.
VR was shrinking. Publishers weren’t touching it. It seemed like even big VR studios were struggling to break even. If we wanted to make a game that sold, VR wasn’t the answer. And on top of that developing for VR was tedious. We quickly realized we’d be way more efficient if we got rid of the need for the headset all together.
Mistake 2: A 3D Puzzle Platformer
It was at this stage that we learned about the wonders of grayboxing. Using basic geometric shapes, usually gray cubes or blocks, to rough out a game world and test a low-fi, simplified version of the environments we wanted to build. This would have been an obvious starting point for any established game studio, but we were still learning how to best approach our prototyping process.
Our efficiency was through the roof. I dove headfirst into reading An Architectural Approach to Level Design and spent weeks blocking out worlds with our team.
We switched to a third person player character and spent a few weeks getting the basic movement and camera systems in place.
We started to feel like it wasn’t enough for the player to just walk around and explore. We weren’t writers. We had ideas for worlds we wanted to build, things we wanted to happen inside them. But the thought of making a game full of dialogue? Most successful walking simulators had rich narratives driven by voiceovers. This just wasn’t possible for us.
We needed gameplay.
Note: Below is a gameplay teaser from our early development phase for Deep Time, the “3D puzzle-platformer” we were making. The video features some of the grayboxed levels we were building out.
After a couple months of prototyping, we had a solid set of mechanics that we could combine into puzzles. We were getting good at blocking out levels. We sent an early demo to some friends and strangers. And people enjoyed it. We had something here.
We started to look at other puzzle games like The Talos Principle and The Witness. We thought: Wow! Here are games full of beautiful environments and solid gameplay that are overflowing with ideas and philosophy. After all, we wanted our game to say something and these were inspiring examples that were huge successes on Steam.
It turns out the team that made The Talos Principle was something like 40 people. And The Witness took a team of industry veterans eight years to make.
Our team? Me, the head of our studio, and a couple of part-time contractors.
What the hell was I thinking?
I went hard on the budgets and spreadsheets for a few days. Could we make this work? Maybe 80 puzzles instead of The Talos Principle’s 120. Maybe a smaller overworld. Not a full open world like The Witness, just something big enough to support our narrative.
In all our excitement, we lost track of the scope. I finally wrapped up the new budget I had been working on, typed in =SUM(), and hit enter.
Two million dollars.
Even after cutting the scope down and reining in our original vision, we would still need an army of 3D modelers to help us create our environments and a couple years of iterating to build out 80 great puzzle levels. This wasn’t going to work. Back to the drawing board.
At this point we started talking to publishers. We had a few meetings with some of the big ones. We cut the games scope drastically and pitched a $400K budget. Enough to support a small team to help us realize the game, but not enough to scare publishers off since we were outsiders and this was our first game. Some were happy to take a meeting, but none were biting.
As the weeks went by, it became clear we had to self-fund and cut the budget even more.
I dove into Steam data, studying a bunch of successful indie puzzle games to try and find something comparable as a model for possible success. I played Manifold Garden, Superliminal, The Stanley Parable, Antichamber, Lingo, Q.U.B.E. Maybe we could make this work with a more minimal art style. We contracted a technical artist for a few months who helped us transform our graybox levels into more art directed environments.
We finished repopulating the 20 puzzle levels with our new assets and started playtesting. People seemed to like the puzzles, they were both challenging and solvable.
Then, one day one of the playtesters said: “I mean, the game is kinda fun, but it looks like a tech demo.”
For context, this was the worst feedback we could’ve gotten at this stage. We had spent months refining our art direction only for it to be called a tech demo, essentially a proof of concept.
He was right. Looking back, we hadn’t really landed on a solid art direction. I mean it looked better than the graybox levels, but it came across as generic. And we didn’t have a unique enough gameplay mechanic that instantly drew players in.
We were operating on a shoestring budget at this point. Our runway was getting shorter and shorter. And we had a tech demo with 20 puzzle levels. And still no positive signs from publishers.
Fuck. Back to the drawing board (again).
Mistake 3: A Procedurally Generated 3D Platformer
At this stage, we’re almost two years deep. We probably could’ve spent another few months adding more puzzles, released the game, and just moved on. But I couldn’t get over the nagging feeling that we’d be releasing a game guaranteed to flop. We couldn’t let all this hard work go to waste! (More on sunk cost in a later post…)
The more we playtested our puzzle game, the more we realized that it was our movement system that drew people in and kept them engaged. Playtesters loved freely jumping, running, and dashing around our levels.
On the side, I had been playing around with making a procedural level generation system in Unreal Engine. It hit me that it could be really interesting to combine this with our movement system from the puzzle game. Every time you died the level would change. The challenge would never be the same. It would be simple, much closer to an arcade game than the immersive, narrative-driven experience we had been imagining.
I took a couple weeks to fine tune a prototype. We sent it around to some of our playtesters. They were hooked.
The feedback was so good we decided to put the puzzle game on hold and quickly pivot. We changed the name of the game.
We spent the next year polishing this. A lot longer than we would’ve liked, but there was so much work to do to make it really playable. And coming from the arts, our standards for the look of the game were high. We didn’t want to just slap together readymade assets like the other games we were playing. We wanted a more unique and crafted look. We brought in a new art director and 3D artist using the last of our remaining funds and completely refreshed our visual direction.
We ended up with something we were really proud of, and we were excited to release it.
But I still had that nagging feeling. 3D platformers just don’t sell well on Steam.
The data doesn’t lie
Here are the stats for the genres we were working in. I use “median copies sold” and “median revenue” as the data points. Median (as opposed to average) is more representative of typical game performance, as it’s less affected by outliers. I’ve also calculated a “hit rate” showing the percentage of games in each genre that have reached 1000 reviews on Steam, a success milestone I covered in my previous post. The data, as always, comes from GameDiscoverCo Plus.
Walking Simulator
Number of Titles on Steam: 3,416
Median Copies Sold: 603
Median Revenue: $588
Hit Rate: 6.85%
Puzzle Platformer
Number of Titles on Steam: 4,485
Median Copies Sold: 240
Median Revenue: $280
Hit Rate: 4.00%
3D Platformer
Number of Titles on Steam: 4,892
Median Copies Sold: 141
Median Revenue: $129
Hit Rate: 3.84%
Okay pretty bad. But how bad is it? Let’s look at some alternate genres that are considered typical top performers on Steam.
City Builder
Number of Titles on Steam: 1,716
Median Copies Sold: 2,013
Median Revenue: $7,701
Hit Rate: 14.82%
Roguelike Deckbuilder
Number of Titles on Steam: 413
Median Copies Sold: 2,597
Median Revenue: $12,129
Hit Rate: 12.59%
Open World Survival Craft
Number of Titles on Steam: 584
Median Copies Sold: 6,588
Median Revenue: $22,184
Hit Rate: 29.74%
Not only do games in these categories make more money, but there are fewer of them. Less competition. Hungrier player base. You’re anywhere from 2x to nearly 8x more likely to succeed in these genres.
Chris Zukowski has a good analogy to illustrate this:
A bad or mediocre product is a rock. When you let it go it falls immediately. Doing extra marketing is almost worthless because no matter how much effort or money you put in, it will fall just as fast.
A good product is like a feather. When you let it go it floats and dances in the air and stays aloft for a long time. You occasionally do some extra marketing which is like giving a gentle puff to the feather sending it sailing back into the air.
Carefully consider your genre, nail your core fantasy, and you're more likely to end up with a feather instead of a rock.
great and honest postmortem. as i keep scarily away from returning to gamedev this helps
Really appreciate your postmortems. It's important to read about the failures as much as the successes.