Three years ago, I landed my first job in mobile game development. I wrote a mystery game. It failed.

I’ll back up a little.

My experience with game writing to that point had been doing creative copy for physical products — card games, trading cards, board games, etc. based on licensed properties, generally — and writing about games, but I really wanted to make the jump into working on digital games. The company I landed at wasn’t a household name, but they had a good track record of mobile releases, were bootstrapped financially, and had a welcoming, collaborative environment, so I was excited to start working on a project there.

I assumed that given my history working on the World of Warcraft Trading Card Game that I’d be placed on an established live game with a fantasy setting, but instead (and to my great pleasure) I was handed an in-development project: a hidden-object game (HOG) with a detective theme. There had been no breakout hit in this genre on mobile; a Facebook game called Criminal Case was the world’s most popular HOG, but the game hadn’t made the jump to iOS/Android, giving us a rare opportunity to lead instead of follow. We understood that it was only a matter of time before Criminal Case hit mobile, so we needed to work fast if we wanted to get the first foothold.

I’d be responsible for all of the game’s content, from coming up with new characters and cases to creating art briefs and photo references for the scenes and items. This was more or less what I was doing for my previous job, plus the added challenge of writing fun, entertaining mysteries.

OK BUT WHAT’S A HOG EXACTLY

(Image: Classic Adventures — The Great Gatsby)

While there are variations in the genre, the core gameplay of a hidden object game is usually as follows:

  1. You’re presented with a scene.
  2. The scene has a list of objects you’re to find.
  3. You tap or click on the objects in the scene to collect them.
  4. A clock ticks down as you play; the faster you find all the objects, the better your score.

There’s usually some sort of story or theme to tie these scenes together. You’ve inherited an old mansion from your grandmother, for example, and finding objects helps you uncover some family secret. Sometimes there’s dialogue; sometimes there isn’t.

If you go to the CD/DVD game section of a Target or Walmart, you’ll see bundles of these games on the cheap. They require very little from the player or their machine. Turns out the HOG genre is hugely popular among women aged 35+, unlike any other project I had ever worked on (or figured I’d work on). Again, a rare opportunity. Women play a lot of casual games, but generally there’s some gender parity or the games skew more towards a male demographic. This isn’t the case with HOGs; the split is somewhere between 70/30 and 80/20. This is why most of them you’ll find feature a female protagonist in the key art or on the cover and tend to feature traditionally “beautiful” scenes. Market research says that’s what the women who play these games want. Criminal Case bucked these trends a bit by having extremely gory crime scenes.

SO WHAT ABOUT YOUR GAME

Our game stuck pretty close to the Criminal Case core loop — that is, what a player experiences during play. It goes something like this:

  1. Begin a new case. This could be understood as a discrete “episode” of a crime procedural.
  2. Unlock a scene.
  3. Spend energy hunting for items in a scene as fast as you can for a high score; in the fiction, you’re hunting for clues to solve a case.
  4. Receive a star when you hit cumulative score thresholds within the scene.
  5. Spend stars on tasks like talking to witnesses or playing “clue analysis” minigames, which advance the story.
  6. Go to 2 as needed.
  7. Go to 3 as needed until you finish the case and confront the criminal.
  8. Go to 1.

Energy is a limited resource, like you’ll find in most free-to-play mobile games. It recharges at a set rate, or you can spend real money on “hard currency” to recharge it to full immediately.

In theory, this system works fine. But we’ll get to that later.

WHAT WAS ALREADY DONE

By the time I joined the team, some initial work had already been done. The game would be a modern-day crime procedural HOG; the protagonist, an ex-cop turned PI who’d been kicked out of the force by a corrupt lieutenant. The team had also set up a cadre of side characters for the protagonist, to act as points of contact for clue analysis:

  • Lindsay, the hacker, who’d handle analysis of technology, like hard drives or security cameras
  • David, the CSI, who’d handle analysis of forensics, like hair or blood evidence
  • James, the coroner, who’d handle analysis of bodies

Ideally, these characters would have unique personalities, adding an additional layer of reward (fun conversation) to the reward players got for spending stars or waiting/paying for timers to expire (progressing the story and unlocking scenes). Art hadn’t been created for any of the characters yet, but we did have visual references and mood boards for each of them, giving the artists something to work with when the time came.

A few scenes had also been created by our art team prior to my arrival for style tests, based on a case concept my predecessor had written up before he left, so we made the decision to include those in one or more cases so as not to waste the manhours used on them. Since I had, essentially, full creative control of all the writing in the cases, I figured I could find a way to include them without much trouble.

We codenamed the game “Crime HOG,” to the delight of pretty much everybody in the office.

WHAT DID YOU LEARN IMMEDIATELY

The company had released one HOG before this, and everyone agreed it really wasn’t very good, but it did give us the most valuable type of code: the kind that’s already done. We’d use that game’s engine (with tweaks for new features and mechanics as needed) and build cases with the versatile “quest” system that all of the company’s games used, from puzzle games to RPGs.

I commandeered one of the many meeting rooms in the office and put a sign on the door christening it “The Crime Lab.” I used color-coded Post-It Notes to plot out the specific tasks and story beats of each case. I learned a few things right away.

Creativity is not a faucet. You cannot just turn it on when you need it. I spent hours staring at the whiteboard, or watching Law & Order hoping for inspiration, or getting mad at myself, because even if you’re creative 100% of your office hours…

Writing mysteries is fucking hard. Even if you’ve read or watched a lot of them, experiencing a particular genre does not necessarily enable you to successfully create works in that genre. Partly due to the type of obsession that only working on “we have to send this to the printers tomorrow or it fucks up our whole fiscal year” types of deadlines in the world of licensed products can engender in one, I was a perfectionist with my cases. I wanted the twists and turns of the case to be reasonable, but not easily predicted. This would have been easier, except…

I had a limited number of assets to work with. Scenes took an extreme amount of manhours to create due to the level of detail required for HOGs (scenes had upwards of 75 different items in them). Also, in an attempt to differentiate our game from Criminal Case, which used flat Flash-game looking character art, we made the (in retrospect) baffling decision to use fully-rendered hand-drawn characters in Crime HOG. These two factors led to a decree: we had to place a hard limit on the amount of assets requested for any given case. Three unique characters, four unique scenes, max, per case. This didn’t seem like a huge deal at first, except…

My narrative was limited by the game’s systems. This is true of any game, of course. But in the case of Crime HOG, it was especially brutal to the narrative. Basically: due to the nature of the game’s core loop, and how it became harder to earn stars (and thus more expensive in terms of energy used) the more stars you earned, players had to have access to the second and third scenes of the case almost immediately (if I remember correctly, after 1 and 3 stars earned, respectively) so as not to blockade them behind long session times early in the case and risk losing them. This meant that there was no possible way to introduce a new scene late in the case as a surprise twist, or even simply keep it out of the way a while. This also might not have been a problem on its own, except…

Limited assets meant limited characters. As mentioned before, I could only add three new characters per case. This seems like plenty on the surface! That means three suspects! But it didn’t really work that way. We needed one character to bring the case to the protagonist, since they were to be a PI. For Case 1 (Garden-Variety Murder), for example, it would be the wife of the deceased, leaving only two characters as suspects. It’s very difficult to craft a satisfying whodunit with two suspects. It could work for the first few cases, until players began to see the seams and understand each case only introduced three new characters. It was definitely possible to overcome this, except…

I was expected to put out cases about twice a month. Obviously this didn’t mean a case went from concept to completion in two weeks — that’s basically impossible. What it did mean was that a case’s art assets had to be in progress (or even completed!) long before I had ever even written the case’s dialogue. So, unless I wanted to go completely insane, I had no option but to simply come up with case “themes” far in advance, and just…write around what I had requested when the time finally came. This might have been feasible, except…

I was alone. I had a content lead who was technically on the project with me, but the studio’s strange hierarchy meant he too was assigned to a game full-time, so the only one doing day-to-day work on the game’s content was me.

SO HOW’D YOU DO IT

Very carefully!

I realized very early on that, mechanically, I would simply not be able to tell whatever stories I wanted. The game’s mechanics weren’t changing, and they severely limited the structure within which the narrative unfolded. So, I adapted. I created a template with the information given to me by the systems designers showing every possible permutation of a case’s beginning. As the case went on and stars were accumulated and spent by the player, the structure requirements eased up, leaving me more room for switching things up and getting creative. The case intro template looked something like this:

Case intro flowchart

This gave between 6 and 7 opportunities for dialogue by the time the third scene was unlocked:

  1. In the introductory dialogue for the case
  2. After the player investigates scene 1
  3. During the first Talk task or after the first Minigame task
  4. After the player investigates scene 2
  5. During the second Talk task or after the second Minigame task
  6. After the player investigates scene 1 or 2 (IF a clue is found)
  7. During the third Talk task or after the third Minigame task

I’ll call these points “conversations” for ease of reference. Conversations could only play during a Talk task or when triggered by completing a quest. Points 2, 4, and 6 and triggered by completing a specific scene play with a specific clue item found — generic scene plays, as repeated by the player to gain stars, could not trigger dialogue since the player was not directed to do so with a task in their case log. “Experts” — Lindsay, James, and David — could not be spoken to with a Talk task. They could only be interacted with through Timer/Clue Analysis tasks (at least initially — more on that later). Conversations were generally around 10 lines of dialogue, varying as the story required.

So let’s break this down.

  • Conversation 1 must be used to either introduce the victim or introduce the client, or a combination of the two, because it’s the only time the player can see dialogue before the first scene is unlocked and played.
  • Conversation 2 must be used to both discuss the clue found on the first scene play. It may also be used to direct players to the next scene, depending on how Conversation 3 is used.
  • Conversation 3 is either used by a Talk task — speaking to a person of interest — or by completing a minigame. Arguably, it was better to use this slot for a minigame, as all of the minigame types involved working with a clue in some way. Just as it wasn’t a good feeling for a player to “find nothing” at a scene for narrative reasons, it also didn’t feel great to find a clue and then not do anything with it. Between Conversations 2 and 3 the player must be directed, through dialogue, to Scene 2.

The process repeated until Scene 3 was unlocked, where I could start making play a little more open-ended, giving the player two or more tasks in their case log to complete at once and therefore letting them receive information in a nonlinear fashion…to a degree. All tasks eventually converged and we could never make tasks a dead end, since players may have spent real money to reach that point. If a player hunted down a lead and it “didn’t go anywhere”, it was actually a required step to unlock another task later down the line that was relevant. After Scene 3 was unlocked, it was also safe to use Timer/Clue Analysis tasks. We opted not to put those early in cases since they could time-lock players from progressing before the first session even ended, which is a pretty shitty experience.

Using the above formula, I wrote 16 cases — 5 for the game’s launch, and 11 more over the short life of the game.

It was super challenging, as were the realities of being beholden to a team and a PM.

WAIT WHAT’S A PM

The role of a PM (project manager/product manager) varies depending on the studio. At mine, a PM was sort of the “owner” of a game project, whether live or in development, accountable to the higher-ups. In theory, a PM guides the project, ensures things are progressing on time, makes major release decisions, and proposes/executes solutions when problems arise. In practice it was a little bit different, with a system that essentially led PMs to micromanage tiny aspects of the game — art, design, etc — for fear of a flop for which they’d be held solely responsible. I’ll provide one scenario out of many.

The time had come in the development process for work to begin on the player avatars. We wanted to offer players a male or female avatar option, as our quest system supported it. I pushed very hard (with support from my lead artist) about providing at least one other ethnicity for the avatar, who would definitely be white by default. Key decisionmakers did not see the value in it, even after we explained how it wouldn’t take nearly as long as coming up with a full new character, so it was set as a “P2.” In startup-speak, a P designation means “priority” — a P0, or Priority Zero, meant the feature/addition/change had to be in the game yesterday. A P1 was a Priority One, an item that was a high priority and might even already have resources allocated to it. P2 was a polite way of saying “this will never happen.” We never openly discussed that “P2” was a feature death sentence, instead cheerily adopting the facade that gosh, well, I hope we can get resources for this one down the line.

The female protagonist came first. Since our demographic was predominantly women, it was vital that her look was approachable and cool, like a TV star, and we’d simply riff on her look for the male avatar. We offered up a series of photo references for her — any given female protagonist on a TV cop show was in there, along with actresses or other celebrities we thought fit the look and feel of the game. Eventually we got a solid look for her.

The male avatar was when things got weird.

We performed the same process as for the female avatar — collected photographic references and sent them to the art team in an email on a Friday, CCing the rest of the game team for feedback and banter. We included some obvious choices, like Cumberbatch in Sherlock, Jonny Lee Miller in Elementary, Matt Bomer in White Collar, etc. Again keeping our older female demo in mind, team members started adding other good-looking celebrities to the pile (“How about RDJ?” “My mom likes a bit of gray in the hair like Mark Harmon”, etc). Eventually some other female team members chimed in saying they were really into Joseph Gordon-Levitt after seeing him in Inception and The Dark Knight Rises.

If you’ve ever been involved in art asset creation for a game, you understand that there are tons and tons of concepts drawn up for any given asset, each using a combination of photo/image references, text descriptions, and the artist’s imagination. Since we weren’t making a licensed game, there was no way any given character would look exactly like one of the many photo references we provided the artists.

This did not stop our PM from getting strangely angry that JGL’s face was added to the pile. I’ll paraphrase.

“That’s stupid. He doesn’t look like a detective. He’s young.”
“Not really; I mean, he’s over 30, and besides, it’s just one reference.”
“We shouldn’t have the protagonist look like him. Don’t use him as a reference.”
“We have 20 other celebrities to use as references; the final version won’t look like JGL. It’s just in there to add flavor and inspire us.”
“We should not just be adding references willy-nilly. This actor does not fit the look the game needs.

From there it escalated, with one artist getting very upset that she was being harangued over a single reference. I talked to my lead and we stepped in, saying “If you have ideas for the player avatar or if you want to provide some data to help the artists, bring it on Monday and we’ll discuss it.”

Monday inevitably came. We figured our PM would simply back off about the issue, but instead we got called into a meeting that took hours. She had rewritten the entire history of the protagonist. “She’s a former CIA codebreaker whose parents were murdered, and she prevented a political assassination” were some key points she wanted to use. We were mystified. The whole conceit of the game was that you were a PI helping to solve street-level crime — bodies found in walk-in freezers, murders in alleyways, whatever — and adding this bizarre new level to it didn’t make any sense. If she saved the President’s life, what is she doing operating a detective agency? If this is her new background, why is there a police CSI and a coroner helping her behind the scenes? These characters worked when she was an ex-cop but seemed seriously out of place when working with what was frankly a superhero. More importantly, this was a mobile game — how the hell were we supposed to convey the breadth of her skills in the limited space provided, in a game about solving murders in a normal American city? And why was a PM stepping in this late in the process to change something like this?

We never really got answers to those questions. What we did get was another presentation. She had data, she said, that proved her point about JGL. Over the weekend, she had set up an online survey with the question “Which actor looks more like a detective?” Then, she put the survey up on Mechanical Turk, an Amazon service that lets you put up menial tasks that people online can complete for spare change. (This is how the “all-emoji Moby Dick” book was written.) She used these three photos of Cumberbatch, Miller, and Bomer:

And used this photo for JGL:

Shockingly, Benedict Cumberbatch won the poll.

All we could really do in response to that particular mess was put our hands up, but we couldn’t change the core conceit of the game on a whim. We managed to talk her down from the CIA codebreaker chess prodigy Navy SEAL sniper she wanted and ended up on the entirely fucking fictitious profession of “consulting detective,” a la Sherlock Holmes. An ex-cop was “too boring,” she said, never mind that we had made the decision not to make the protagonist a cop specifically to free us from legal constraints in the fiction. A connected ex-cop could do more than a cop could to help people while bending or even breaking the law, and it’d lead to more interesting cases. Now we were left with what could only be described as “a smart person” as the protagonist, all other details neutered. I had written a few cases already that would need to change. We now had a game and cast built for an ex-cop that did not feature an ex-cop.

The PM left the company prior to the game’s launch. Her changes remained.

But we did launch the game. Over the 2013 holiday season it hit #4 on the iTunes Store in Games, stayed on the top ten for a while, and then began a slow, agonizing descent into oblivion as I watched.

SO HOW’D YOU PERSONALLY FUCK UP?

Let me count the ways!

No bodies. In early meetings we all sort of bristled at the level of gore in Criminal Case scenes — mangled bodies in dirty bathrooms, people impaled on meathooks, etc — and decided to go a little more PG. The idea was, people grossed out by that type of scene would keep playing the game instead of leaving, and people who weren’t would never miss them anyway. None of the scenes in the launch game had bodies in them, either, but this was largely due to not knowing the story of their respective cases at the time we ordered the scenes. We justified continuing to do it with the “PG approach” mentioned above. I still think CC’s scenes are a little much for a mass-market game in an older female demo, but we were dead wrong about not including bodies in the scenes. Bodies not only helped set up cool tableaus in scenes, but gave the player an immediate, visceral connection to the case. They helped make the mystery worth solving — if you never saw the victim you couldn’t be asked to care about the case’s result. I rectified this by case 6, adding body parts in a scene, and finally a full body in case 7 and each case after that.

No partners. This was a big one. Early on I decided I wanted the protagonist to be a unique character, not just The Player. I think this was a great decision that led to fun dialogue. But, it introduced a huge problem I couldn’t immediately fix. When she found a clue, when she opened a new scene, when she was done talking to a client or suspect, the protagonist was constantly fucking talking to herself. By relegating all of her posse to their analyst roles, I had given her no one to talk to, to banter with, or decompress with out in the field. And I didn’t want to give the game a narrator, opting to let the protagonist’s voice be the voice of the game. But I also felt like I couldn’t just instantly drop a new character in there as her partner, feeling I’d be doing the players who gave a shit about the story a disservice. Some other content designers at the company thought I was crazy, because it was just a mobile game and nobody cared, but I figured if I showed I gave a shit, players might too. My solution was to introduce Miranda, a cop, in an earlier case and eventually having her come work with the protagonist full-time in a later case following some drama.

Not all of the cases were murder cases. This was a stupid early decision on my part. Playing into the “PG approach,” I thought I could make some cases that weren’t homicides to “spice things up.” Murder all the time would lead to fatigue, I thought! No. There’s a reason why Law & Order’s formula is so successful: it works and people like it, and murder mysteries are easier to write.

I didn’t know my demo well enough. The first few cases were pretty on-point when it came to providing our target audience with what they generally liked, but case 5 was a stinker in pretty much every way. I had ordered scenes for a sports locker room, a dugout, and a baseball stadium. When I ordered them, I had absolutely no idea what I was going to do with them — I just knew they had a cohesive theme and I could figure it out later. Well, I couldn’t. Instead of making it a murder case, I had the detective come in and try to figure out whether a star baseball player was throwing games, by the coach’s request. Not only was the scenario not traditionally interesting for our demo, I didn’t know shit about baseball, and I certainly couldn’t figure out how to make an exciting mystery out of sports gambling. It was the only case I felt really, really bad about when I handed it in.

Not enough attention to the game’s serial nature. All the early cases were more or less episodic, which works if you land on TNT when you’re flipping through channels, but if you were on case 9 in our game, you had already played through 8 other cases. I didn’t make enough of an effort to keep storylines going. This was due in part to the nature of how our cases were structured and the limited number of characters I could add. If one character of three is the culprit, that leaves two characters I could use in a future case, but were any of them even interesting enough to recur? I tried to start doing it very late, by having a sequel of sorts to Case 2. It was one of the game’s last and it was very silly, but it felt nice to have done it, at least.

Bland side characters. I worried about this one early on. I got tired of gimmicky side characters in games really quickly, and I didn’t want them in my game. The friendly character that players responded to the most was Lindsay, who had cool hair and tattoos and a fun attitude and, probably most importantly, was a woman. David and James were boring as hell early on. I didn’t know what to do with them. James, the coroner, had one of the goofiest pieces of character art I’d ever seen in a game like this, with a hotdog in one hand, a glass eye, and a host of other over-the-top details, leaving me with (what I felt was) essentially nothing to work with in characterization. I thought using the old “weird coroner” characterization they use on practically every procedural out there was rote, but also didn’t have a competent plan in place for him. In a way, David got an even worse rap; he was the sole black member of the main cast, and arguably the sole person of color in the main cast. I was unequipped to make him Black, I told myself, and thus I overcorrected into giving him no personality at all for fear of writing a stereotype. It was a cop-out. I tried to fix both characters later by giving them background details they could slip into conversations, giving them more room to grow. Unfortunately, players still had to deal with them for the first however-many cases of the game. They were a huge missed opportunity.

SO ALL OF THAT STUFF CAUSED THE GAME TO FAIL, THEN

Not really, no. I consider them personal and narrative failures, but they weren’t what killed the game. Basically, the game shipped without any way for a player to replay old content. I don’t necessarily mean replaying old cases from beginning to end; I don’t know if that’d be fun. But a tremendous amount of cost and effort went into each case’s art assets, and writing cases took quite a bit of time too. But then when a player beat a case, those assets were never used again. When we had originally scoped the game, we had two features to help with replay value and increasing ROI on case assets. One was Detective Skills, which in RPG terms was a sort of “talent pane,” where players could spend excess stars for permanent bonuses, like extra time in timed modes or extra points for finding furniture objects or what have you. Another was the Trophy Rack, which would give you rewards for getting all five stars in any given scene. The argument was that without a reason, players wouldn’t replay old content. Not only were the features cut for the MVP (minimum viable product, a startup term for the piece of shit version of a game you launch with), players literally couldn’t even go back to a previous case even if they wanted to.

While the game enjoyed a lot of popularity at first, thanks to some marketing magic I still don’t really understand, within the first few weeks we knew we weren’t going to make a profit on the costs of producing scenes and characters. Obviously we recommended adding the features described above, but the studio had a weird approach — if the game did well, it got engineering support, but if it did poorly, then it didn’t. The problem there was, of course, that a game generally needed engineering support for new features or changes in order to do well. Eventually we got the smallest amount of engineering hours to put in a new feature the design team barely even knew about: PvP Challenges, where the player was ostensibly matched against another player to see who could get the highest score in a scene. It could draw from all scenes in the game, making it a cost-efficient feature in that regard. However, players were never actually paired with anyone. The game just provided a random number for the other “player,” within a particular range. It was a farce. The only exciting part of the feature was that completing challenges gave you a dog that would, on occasion, give you a higher score. You never got to interact with the dog, or see it in scenes — you just uh, had the dog, somewhere.

The feature stank and we all knew it, even the players. The fact that it didn’t work to raise metrics in any meaningful way meant that the higher-ups were Right to Not Give The Game Support, as it were, so away it went.

WHAT DID YOU DO RIGHT

The game was funny. I had to make it funny, because mobile games are a dime a dozen, and nobody remembers a mobile game for being serious. I figured if I could make a player laugh during a session it’d increase the odds they’d come back for the next. I got genuine laughs from the team and company during group playtests and it made me feel really, really good. Posters on our company forums would point out jokes they liked, too.

The protagonist was queer. This was sort of an accident at first. See, while we could offer a different avatar based on players’ choice, I couldn’t actually change the text to match. So whether the male or female avatar was chosen, conversations would always play out the same way. In a way I had sort of forgotten the male avatar existed since I always played as the female one in playtests, so when it came to David hitting on you, or Lindsay flirting with you, they didn’t care about your gender and neither did you. After noticing it in a playtest I just decided to roll with it and just had everybody hitting on everybody willy-nilly. And this is a small thing, but I was really proud that the first thing the player reads, the text above the two avatar options, is “Which detective would you like to be?” instead of the “Are you a man or a woman?” Message that was originally planned. Between that, the fluid sexuality of the protagonist, and the fact that pronouns were never used for the protagonist, I like to think that I made a game where a queer or trans or even non-binary person could feel welcome, within my power.

I made some good Frankensteins. Remember the three scenes that were already made before I joined the team? Eventually those caught up to me and I had to write a case with a barber shop, a pawn shop, and a projection room as its three scenes. Somehow I managed to write a good mystery out of it, even if it broke my murders-only rule by focusing on an apparent theft.

I actually wrote that shit. It was a pain in the ass and it stressed me out but I actually did it. It was a miserable failure and I wouldn’t want to do it again, but I did it, and it was all me.

Also, somebody on Tumblr shipped the detective and Lindsay. That’s when you know you’ve arrived.