If you were to sum up Zach Arundel’s job in two words, those words would be "being thorough". As a data scientist, he spends his time counting, re-counting, then analysing and evaluating huge amounts of data about Dreams and the Dreamiverse. It's time well spent though, helping inform which game features Mm works on, and how we respond to coMmunity needs. Here, we chat to him about his first internet search, impostor syndrome, and his stint definitely working for MI5. Maybe.
Hi Zach! What do you do here at Media Molecule?
I'm a data scientist at Mm, which is kind of a slightly newly made-up job title stemming from the late 2000s, when data started to become a thing people actually cared about. Basically I do lots of counting and analysis of our players' behaviour, and it always boils down to working out how many times something happened - for example, how many times players jump in a level, or how many collectibles they'll find before they stop trying. So we do lots of analysis of gameplay and user interactions for our big Mm releases, a lot of work with the curation team, a lot of AB testing and randomised control trials. It helps to inform what we work on next, and to respond to player and coMmunity feedback.
How does that data get used once you've collected it?
Well, this is probably the most fun part of the job, where you get to talk to the designers and the team about the game they released, and basically do a post-mortem of what was successful and what didn't work. Using this information, we can figure out how to build on our successes for whatever comes next, and find data-driven ways to make our creations more appealing to players. Obviously we do a lot of user research, as without great user research, you don't necessarily know how players and creators are responding to the additions and updates we put in the game. That said, our design teams are very experienced and know their audience, so they already have pretty good intuition about the friction points and more challenging aspects of our games that might be a turn-off to certain players.
Do you regularly perform analysis to see what those sticking points are for players?
Yeah, so we did a bunch of this on Ancient Dangers: A Bat's Tale, which was quite early on in my Mm career. It was one of those rewarding things where I looked at the player progress through Ancient Dangers and talked with the designers about how we could smooth some of the rough edges. In almost any other game, this'd be quite a standard process, but Dreams is an interesting platform in that anything you make on it always has this direct competition of every other creation in the Dreamiverse, and therefore this indirect competition in the sense that there might be something better to be playing. And they’re all free, so there's no barrier to quitting out and finding something else instead. So you've got to be a bit more open and generous with players, and perhaps not treat it like, 'Hey, you just dropped £50 on a new game, so I'm gonna cutscene you right up, and then I'm gonna make it really difficult'. I guess the most interesting aspect was that the things we encountered with Ancient Dangers were all issues recognised by the designers, but they had kind of forgotten, you know, somewhere in the dev process.
But game development is just so hard and things move so fast. And a new problem always arises, so you just have to keep moving on, or it can hold up other parts of development. But the nice thing about data and user research is you get to see stuff with fresh eyes that you couldn't see when you were making it. By the time you released the game, I don't think you can even see where the problems are any more, as it will have changed so much from when it started several years ago. All you see is this thing you just poured so much into, and I think you can only see it again when you're talking to user researchers or when you're talking to analytics and they give you another lens. And then when you see how people actually play the game and respond to it, you're like, 'Ah, yes perhaps that could be changed'.
So the most fun part of the job is talking to the designers and giving them information so they're redesigning the game in the room with you - it feels really rewarding, and like you're really participating in improving the gameplay experience for users. But all the hard work and tweaking pays off, as we were able to release a later update for Ancient Dangers that smoothed out a lot of the issues that we had picked up from players when conducting user research and analysis.
It's interesting you mentioned cutscenes as a potential barrier for players. Data-wise, do long cutscenes at the start of a creation put players off?
It does have a big impact. I wouldn't want to put anyone off from doing them - like, if your vision is for a really grand opening, you should do it - but it might cost you players who aren't so interested in a narrative experience. It's not an issue if that isn't what they wanted to play anyway, but they will never know if they would like your creations because they weren't ready to receive what you were trying to sell them, and they may have been more receptive to it later on. That's maybe a bit philistine-y, but Dreams is a very tempting place. It’s like being a kid in a sweet shop. If you’re a sweet retailer, you don't hide your nice sweets behind, like, a challenging artsy wrapper and five minutes of sweet lore - you make it pop so the kids immediately run over to rot their teeth on them. And then maybe you give someone a nice bit of story or a plot dump after they’ve got a taste for the game, as a treat.
Is there any really weird or unexpected data that you've discovered about Dreams?
I was looking at the most popular searches, and Dreams players are specifically interested in giant dinosaur creatures. Like, kaiju-sized angry monsters. I wonder what the venn diagram of dinosaur fans and the Mm community would look like. The crossover of Sackboy and a bunch of realistically rendered dinosaurs seems to be a really substantial one that I had not anticipated. I mean, obviously I like smashing stuff, and dinosaurs look cool, but the Dreamiverse is swarming with them.
My theory is that exploring the Dreamiverse is like the first time you went on the internet. I remember being at school at lunch time and they just sat you in front of the computer and let you go wild on the internet (with the safe search on, obvs). I'd never been on it before, so I was just looking at my mate and copying him, typing in bbc.com. The BBC website must be one of the first websites that, like, 90% of people in the UK go on other than, like, Google. Then after that, the next time I typed something stupid like sausages.com. I think people have that situation in Dreams where they just see a search bar and type in whatever pops into their heads. For a lot of people it seems to be cool stuff like dinosaurs, but for me I guess I just like sausages.
How did you get to Media Molecule, then? Was data science something you had worked in for a while?
I actually used to work for the UK Government, and I was a data scientist there too. But I was just doing various boring government jobs that I progressively disliked more and more. Over time, there was this growing interest in data science within the government, so it was something that I explored getting into. I didn’t do that for too long, but since I had gained the skills for data science whilst working there, I thought maybe I could have a crack at my dream job, working at a game studio. So I applied to Mm, and here I am!
Alright, level with us. Were you a spy?
Everyone thinks I was in MI5. And the more boring I make my job sound, the more they are convinced it was MI5, but no. I did work in government ministers offices for a bit and that was super fun for all the good gossip - it was like an episode of The Thick Of It. But one day I had a really bad day at my old job, and I told myself, OK, I gotta look for games jobs now. Luckily Mm were recruiting, and my dream came true.
Fine, but we're always gonna be a little suspicious. Did you study data science at all?
Actually I did a chemistry degree at university very, very poorly, so my transition to data science was completely unrelated to my degree. As a result, I’m self-taught in data science, but I was lucky enough to get that first job, and then learned everything else on the job. I think all jobs have some level of impostor syndrome, but data science is really bad for it because you're often working with really smart people and you're very conscious that you're not nearly as intelligent as them. But you just have to find a team that accepts the variations in areas of expertise.
I used to work on a team with someone who was so good at statistics that I had to re-evaluate everything I thought I knew about numbers. But it was a great working relationship, as they could do all the stats and teach me how to improve, and I handled the engineering side.
What kind of tools and analytics do you use day-to-day?
So we work with our programmers to put in telemetry hooks that can report interesting player data when something happens. And then that goes into a big old database, and I write loads of SQL queries all day. So for anyone who's thinking about becoming a data scientist, just learn how to be good at SQL and you are 80% there. And if that seems intolerably boring, definitely look elsewhere. I happen to like writing SQL.
You learn SQL, you create the database, and then you run your Python scripts which make the magic happen. There's a lot of farting around with data and that's the main part of the job, tidying it up and making sure it makes sense and removing all the 'outliers'. Then you present it to the relevant people, and make sure they understand it so they can act on it and improve the game.
Do you have any other advice for people that are interested in getting into data science?
Yeah, the learning on the job thing I mentioned earlier is incredibly important. I had a really great mentor at one of my jobs. And I was a bit like a vampire just sucking up all their knowledge. And everyone says this in these interviews, but: show that you are passionate about whatever area you're going into. People will dig that. It really helps to be interested in what you’re working on, as I would be 90% worse at my job if I wasn't working on Dreams. But I love Dreams and I love everyone that plays Dreams, and I'm interested in what they want to do. In my previous jobs, I didn't really care what people used the websites for, you know, because it didn't mean anything to me. But video games and Dreams mean so much to me that I'm really interested in what people are doing, and I really want to make it better. So find something that you're actually interested in, and care about the users, and you will be a better data scientist for it.
Mm staff seem to have a lot of interesting items on their desks. What's on your desk at the moment?
The state of my desk is a point of repeated contention with my wife because I eat lots of clementines, especially at this time of year. But I also do like to eat the orange peel, which you have to eat slowly, because it's so bitter. As a result, my desk is always covered in oranges at different stages of being consumed. It does leave quite a pleasant orangey fragrance, but is definitely messy.
There are some books that are strategically placed, because this is the guest bedroom as well, or not the guest bedroom if the guest is my infant son. But the books are there to give the impression that a smart person might work in this room occasionally.
To wrap up, what are your favourite Dreams, and are there any that you'd like to recommend?
I wanted to pick ones that don’t have a million eyes on them already, which ruled out a lot of my absolute faves, but they all got big love already, so never mind. So DEAD DREAMS 2 (by SmileBasic) is one of them. It's basically a collection of all these half-finished creations and it’s so generous because it's got, like, 29 things in it, but they're all totally good enough to have just been bashed out anyway. There's this driving game where you platform around and a funny football game. It has that slightly weird, minimalistic, homebrew vibe. I love generous Dreams, where you just feel like they didn't need to give us so much but they did anyway!
Number two is Defender Imp!, which is an absolute modern classic shoot-em-up. I wish there were more games in Dreams like this, as it’s just such an economical design, so it's kind of perfect for Dreams. I know it's a bit hot right now, but I just love it when any creator does a design that makes it easy for themselves. Easy to debug, easy to test, easy to play - like, you don't need to make millions of beautiful assets. It just feels very smart, perfectly executed, and I love a lot of these light and breezy R&D style designs. And that's a perfect example of my favourite creation style, so well done Sharfik1995.
My final recommendation is Grapple Swing Platformer by Agent_C0w161. This creation is from 2021, and it's got quite a restricted grapple hook mechanic, where you can only anchor on fixed points - but the swinging is so cinematic yet also quite freeform, and it makes it really satisfying. It's not super complex, you're only playing as a little sort of cube person. But the way you elegantly swing around this geometry, coupled with the cinematic quality, makes you feel a bit like a cuboid Nathan Drake flinging yourself around with his grappling hook even though you're this very static cube.
It's not really tightly stage managed like Uncharted, so it feels even better because you look cool and you’re pulling it off yourself. It all flows really naturally. It's like I had to press the button and I had to aim, instead of the game doing it for me. So Grapple Swing Platformer is my personal hidden gem. That’s if something that has been played over 500 times can still be considered a hidden gem!
The Dreams User Guide is a work-in-progress. Keep an eye out for updates as we add more learning resources and articles over time.