GDC 2018: Yokozuna, big game data, and the future of MMO monetization

GDC isn’t E3. It isn’t PAX. It’s not even what I think stereotypical gamers can appreciate. But I think the Massively OP crowd is a different sort, and because of that, we can give you some content the other guys might not be talking to you about. Like data collection and monetization. They’re necessary evils, in that we armchair devs can generally see past mistakes rolled out again, but know those choices are being made in the pursuit of money.

So how do you make better games and money? Maybe try hiring some data scientists, not just to help with product testing and surveys, but with some awesome, AI-driven, deep learning tools. Like from Yokozuna Data, whose platform predicts individual player behavior. I was lucky enough to sit down with not only Design and Communication Lead Vitor Santos but Chief Data Scientist África Periáñez, whose research on churn prediction inspired me to contact the company about our interview in the first place!

What does Yokozuna Data actually do?

Without getting technical (because I can’t!), I can at least convey that the system uses machine learning and artificial intelligence to learn more about gamers and the game market by combining with deep learning (a kind of AI that tries to imitate the human process of decision-making). This makes it so the program is quite adaptable in both terms of not just the game platforms it can be used on (mobile, console, PC), but the genres as well (puzzler, RPG, MMO, etc).

Developers using Yokozuna need to first upload the data and let the program run for a week. After that, the developers have access to a dashboard that gives them the information on their users. It’s not just about tracking either, but tools that can help both the customer and the player. For example, there’s AI recommendations to improve the game and predict how players will react to potential changes.

See, while lots of games perform well, developers and management don’t always understand why. Machines may not understand things the same way as people, but they’re great at going through a lot of data, and the data can help humans better pinpoint how that success came to be (and ideally, replicate it). Yokozuna’s model just takes that a bit farther than what people might traditionally be used to.

For example, previously, these models grouped players together with people who seemed similar to them, but Yokozuna can track and predict individual player habits separately. I’ll get to some of the implications of that later, but as an example, this can help a company understand who the player is. Based on when the player is active, we can figure out whether the players are students or salaried workers and whether they’re playing at home or while commuting.

From left to right: Vitor Santos, Paul Bertens, África Periáñez

Using the data

While Dr. Periáñez now sadly lacks the time to play games, she used to be a big Tomb Raider fan. Research not only took up her time, but also got her to notice design issues more and more. For example, a game might spam you with notices that clearly don’t apply to you. Even if you don’t play mobile games, we’ve all had mobile apps ask us to come back to them mere minutes after closing them or been told about sales after recently taking advantage of said deal. It comes across as spammy and ill-conceived.

Yokozuna’s system is supposed to be better than that. Since the system can track individual users, it can be used to recommend items to players, and not just any items. The system can predict what users are probably aiming for, and companies can even use the tool to offer a discount to the player. We’ve all had our “I quit!” moments fizzle out when we finally get that drop, or a sale pops up that lets us cheaply buy something that makes our goals more achievable. Having companies know when our “I quit” moment is coming and gives them a way to keep us can be beneficial for everyone – though it can also seem uncomfortably manipulative, like the sort of thing casinos do to keep you in your seat and feeding money into a machine, which we’ll get into later in this piece.

The data aren’t just limited to individuals, though. We all know that “one company” that is deep in a foreign Chinese community, ignored Lunar New Year their first year, and then held an in-game event for just two days. A quick Wikipedia trip would have told the clearly community ignorant developers that not only is the holiday, on average, celebrated for three days, but it’s a time of year when kids are getting the US equivalent of hundreds of dollars. It’s like freaking Christmas, and not hosting a cool event (let alone nice deals) for more than a few days is literally throwing money away.

For these companies, Yokozuna can simulate events to see how it affects sales and engagement time, not just globally, but regionally. As Dr. Periáñez notes, an event in Japan may not get the same results as a game elsewhere. If these certain AAA companies refuse to spend money on community managers who “get” their audience and can share cultural information, they should at least consider getting AI to show them the data on why it might be a good idea to cater to their community.

And that’s the key for them at the moment. Right now, Yokozuna is trying to educate developers and businesses how this tool can be useful to them. Dr. Periáñez feels that game data and the industry is very conservative, and she’s worked in data science for a few different fields. She says games have some of the most detailed information on human behavior since so much of it can be tracked in-game and can be followed up on, especially since games make certain habits easy to predict. In fact, Dr. Periáñez feels game companies are behind energy, communication, finance in terms of utilizing data research, and my own experience (living in two countries, my family from a third, working in blogging, education, and local government) gives me the same feeling.

That’s why they’re teaming up with Unity, Ubisoft, DeepMind, and others for a summer school session on AI and games for university students and professionals. For readers who are also frustrated with the current state of data research in games, this may be an opportunity for you to make a deep impact on our hobby.

Potential for abuse and misuse

While some of what I’ve said may come off as extremely optimistic, let me say this: Some of what I saw and heard makes me a little concerned. I’d be lying if I didn’t say that I had moments reminding me of sci-fi mad scientists, wielding awesome power but pushing ahead without considering all the consequences.

Much of the data is collected because you, the player, actually agree to it by accepting EULAs. This is what allows Yokozuna to track your history of actions, purchases, even who your friends are, not just to see who may be the one most engaged with the game but who’s a leader of your group and who might provoke others to leave. This is what gives Yokozuna the power to predict when you think you might quit, what it thinks you’re trying to buy, and when to offer you up that pixelated treasure to keep you playing for longer.

On the one hand, that’s Yokozuna’s job and the job of all companies: to get you to part with your cash. However, this is also how, perhaps unbeknownst to the companies, they could be tempting a gamer with problematic habits. You can call it addiction if you want, but we’ve all had times when we chose to play a game a little longer than we should have, or spent a little more money than we really wanted to. This kind of data-driven psychological manipulation is one of the things people use to compare our hobby with gambling right now, and it’s exactly the sort of system that Activision has patented, to the outrage of gamers.

And that’s another thing that’s more than a little scary. I wasn’t terribly surprised when Dr. Periáñez says that Yokozuna could still do research to help monetize casinos, despite the fact that her game data research is vastly different from what casinos do. Casinos have been doing their research for far longer than the video game industry, and they have (among other things) different users and motivations, but there’s enough common ground to bring up concern for me, a big video game enthusiast that really doesn’t want my hobby to fall into the same restricted category gambling does.

Another alarm sounding off right now? Privacy concerns, the use and sharing of our data between companies, and of course, giving even singular companies the power to personally tempt us to keep with their product when we may be struggling to break free. Personal responsibility is important, but that battle gets harder when you’re fighting data-backed manipulation specifically crafted to keep you engaged.

That being said, there is some good news. According to the company, no one’s yet approached Yokozuna about making a game to get kids to gamble or encourage harmful habits. In fact, companies (especially in Japan) that rely on mechanics that resemble gambling, like gachapon, have asked for other methods to get people to spend money. Yokozuna’s suggestion is often to pair it with events, like holidays, social events, raids, etc. It may be lootboxes, but it may also be a specific item to make an event easier.

Yokozuna is also looking at how to use its tech with health. I can’t say for certain that its data won’t be used for corporate evil, but I felt like the intent from Yokozuna (and its current customers) is to make games engaging as well as profitable, not simply to milk your wallet. If the data can also get people to do things that are better for them, that’d be awesome too.

What the data say about potential MMO monetization

As usual, I’ll pose the question some of you already are asking: So what? This is an MMO site. What does this have to say about our genre?

Monetization is king. Of course companies want to get as much money from you as possible. However, as Santos mentions, it’s in everyone’s best interest to have happy players, period, even if they don’t pay. According to their research, whales may often only make up 1-2% of a freemium game population, but account for about 50% of the purchasing revenue. Whales also tend to last in-game the longest.

This means that developers who aren’t looking at the big picture may only focus on whales as they keep their games afloat. The problem is that said whales may also be engaged because they’re the center of their social circles, and if a whale’s free to play friends jump ship, it’s entirely possible that person may jump ship too. It’s something Yokozuna could help a company predict and prepare for.

Somewhat related is that buy-to-play models are vastly different from freemium in that people stick longer initially due to a bigger investment. However, smaller purchases are usually more engaging than large ones, which is why there’s a difference in audience, with subscription models somewhat in between but also enforcing a hard start/stop engagement period. People might spend the same amount of money or more if it’s smaller. I’ll let you all fight about which one is the best down in the comments, but we all know that, as usual, these data confirm anecdotal arguments we’ve been making for years.

Again, monetization and player retention is very relevant to both our MMO interests and what Yokozuna does. For example, the team won not one, but two awards in predicting two tracks of research related to Blade & Soul player loss/retention. Normally that’d be kind of cool except for an interesting twist: The data were collected during a time the game moved from a subscription model to a freemium model. It means the team’s research is quite flexible in terms of models, adaptability, and doing what it sets out to do: help keep people playing (and the studio afloat).

As Dr. Periáñez notes, game companies are often chasing the same players and the same markets. Ninety percent of free to play app users stop playing after the first day, and 5% remain after three days. As we often see in the comments section, MMOs aren’t engaging our MMO audience on mobile quite like PC can (and to an extent, console). Yokozuna doesn’t just want to try to help developers make games better; it’s specifically looking to help with player retention and to prevent customer churn (the technical term for subscriber/customer loss for anyone who wants to dive into some research).

We’ve all seen how bad developers can go so wrong without proper metrics. To see devs run on their (incorrect) gut feelings over and over, in spite of data to the contrary, is actually kind of frustrating. For a fan, it sucks to be disappointed, but as industry, it sucks even more to know that we’ll be writing about yet another MMO predictably making all the same avoidable mistakes for years. Data research can keep them on track – if it’s used ethically and responsibly. And that’s the next hurdle for the industry.

Massively Overpowered was on the ground in San Francisco for GDC 2018, bringing you expert MMO coverage on everything (and everyone!) on display at the latest Game Developers Conference!
SHARE THIS ARTICLE
Code of Conduct | Edit Your Profile | Commenting FAQ | Badge Reclamation | Badge Key

17
LEAVE A COMMENT

Please Login to comment
  Subscribe  
newest oldest most liked
Subscribe to:
OverdriveActive
Reader
OverdriveActive

I think this outcome is kind of expected but definitely a world away from where we used to be. If you’ve been playing games for a few decades this will definitely feel unnatural. I think we always wanted developers to just make ‘fun’ stuff and then the consumers would flock to those games. With gaming becoming much more popular and mainstream, hundreds of millions being spent on development and employing hundreds of thousands of people globally it’s not really surprising that they’re receiving the same consideration that McDonald’s colour design gets. They pick their setups and colours to get people in and out the door so they can serve more people, MMOs get analyzed and reconstructed to drive purchases and try to get you into an addictive behavioural loop.

For people entering gaming now this will all probably look very different. The threshold for being feeling taken advantage of or being squeezed will be much higher. The way they will engage with games as a whole will probably change to match as well.

I think anything else I will say here will be speculation so I’ll just leave it at that. I’d just like to say keep in mind that big games aren’t made for what you might think are “gamers”.

In my head the word “gamers” still evokes visions of classic nerds but I think that’s no longer accurate. The D&D, Warhammer, Extended LOTR DVD release viewing party on IRC variety is where I came from. With hundreds of millions of dollars on the line, you better target the Instagram, Twitch cult of personality, grown up Minecraft player instead. Parting them with their money will be much easier. The nerds will scream and yell but they aren’t the spenders anyway. Let them stew and make your games for the new generation of gamers that doesn’t know any different than microtransactions and lootboxes. As a developer and publisher you’ll be better off.

I don’t think that’s a jaded view either. I think that people like me and people of my type of variety just need to direct their attention more to making sure games that do cater to us are still possible 10 years from now. The way MMOs are going we’re going to have a hard time getting those games made at all. Maybe Kickstarter / IndieGoGo funding rounds are a thing that helps do this.

Reader
Patreon Donor
Kickstarter Donor
Loyal Patron
Stormwaltz

Ya’ll need to look at Daniel Cook’s presentation on “Designing Friendship,” because it articulately analyzes things I’ve inarticulately felt for years.

Here’s a link to his powerpoint:

https://docs.google.com/presentation/d/1GXoKH5ltEZDBXf-F6i2gIjY03fQKYQDRLhxOr3JzES4/edit#slide=id.p3

OverdriveActive
Reader
OverdriveActive

Thanks for this, it was a good read.

Reader
Patreon Donor
Kickstarter Donor
Loyal Patron
Stormwaltz

The full video of his presentation is online now for free!

https://www.gdcvault.com/play/1024955/Game-Design-Patterns-for-Building

Reader
Arktouros

This is all a huge reason why I say we can’t go back to the way things were with a subscription model. Game companies into data analytics are too informed and understand that a system that allows them to tweak their business model as they go is too powerful of an economic tool.

While people like to think they’re immune to this, a lot of times this is not done overtly. They aren’t going to reach out to individually sell you a particular item. It’ll be done at a trend level such as after X days we’ll notice Y drop off that selling Z item type will have such and such impact on retention. Claiming you’re going to quit if you see that is perfectly fine, acceptable and even planned for with these metrics. They know you’re going to quit. They’re okay with that.

Once you understand and fully grasp what this means what you do or don’t do doesn’t matter. Ultimately they have your data recorded and will be able to predict what you’re going to do before you decide to actually do it. Another big reason why I preach stop worrying about all this nonsense and just have fun. There’s really nothing you can do about it, might as well just have a good time.

Reader
Toy Clown

A big thumb’s up for the article. Thank you.

Reading this leaves me feeling disturbed. Many game companies are already crossing lines delving into psychological aspects to part us with our money, such as creating addictive spending habits such as lock boxes and RNG loot boxes, using the carrot dangling on a stick and yanking it back routine, offering cash investment vs. time investment, etc.

The whole … thing, being between gamers and developers is getting sketchier. So many players refuse to pay for 24/7 access to games, but are easily tricked into paying exorbitant amounts of money on psychological yank strings created to extract money from those unwilling to pay for their product.

This sort of data reminds me of what I used to do when I was in Uni, where we got paid 20-100$ to be a test subject for questionnaires, impulse testing, medical testing and other tests. (I earned money by doing this and selling my plasma while I did uni!)

That leads me into another thing that bothers me is that this data collection is done for “free” and we know it will end up in the hands of game developers as use to trick us out of spending more money. I miss the days when costs were straight-up-on-the-table and choices were left in the buyer’s hands — not what they can be tricked out of.

Reader
Schmidt.Capela

We’ve all had our “I quit!” moments fizzle out when we finally get that drop, or a sale pops up that lets us cheaply buy something that makes our goals more achievable. Having companies know when our “I quit” moment is coming and gives them a way to keep us can be beneficial for everyone

I might be the exception rather than the rule, but those examples would have the opposite effect on me.

If I ever get the impression the game is tweaking drop chances behind my back in an attempt to keep me playing I’m likely to leave on the spot. Besides, I hate RNG drops enough that I make a conscious effort to forget that kind of drop even exists, so if the item is only obtained as a random drop I will likely be ignoring its existence, and thus not going after it at all.

As for sales of items that would help me achieve whatever I have in mind, those would just serve to remind me that the devs put more importance in fleecing the player out of every penny than in actually balancing the game to be enjoyable, making me more likely to quit the game instead.

Veldan
Reader
Patreon Donor
Veldan

I may be another exception, but I’ve quit more times because of getting something than because of not getting something. When I finally get what I’ve been chasing I sometimes feel “done” and can’t be bothered to log in again.

Reader
Alex Malone

I like these sorts of articles so good work on getting the interview and passing on the info.

As for the substance, this sort of data analytics is a very tricky area. Getting good data is critical for making business decisions so the more data a business is collecting, the better.

However, the data is only the start. Data collected is all about what has already happened and can only be used to predict what your existing playerbase is doing. That means making decisions using only data is unlikely to be helpful. It can’t help developers come up with new ideas, it will only tell them what existing ideas have worked.

I’m also curious how MMO developers are going to deal with the GDPR (general data protection regulation) that comes into force on May 25th. I work in IT so we’ve been focusing on the regulations a lot recently and from May 25th onwards, we as data subjects are getting loads of additional rights.

So, in this scenario, the developer sharing data with a company like Yokozuna is not essential for their game / service, so their legal basis for sharing is “consent”. That means we have to manually consent to them sharing data. I also believe that because the sharing is not essential, we cannot be blocked from accessing the service if we don’t consent.

I know that I will personally be putting these new rights into action and I’m really curious what the results will be, but I’m also curious what effects the GDPR will have on companies like Yokozuna.

Reader
Brother Maynard

I’m also curious what effects the GDPR will have on companies like Yokozuna.

I think they will opt to act at the layer above the data collection. They will simply ask the companies hiring Yokozuna’s services to certify that the data they are requested to process have been obtained and are used in accordance with the legal requirements.

Edit: I did word it a bit vaguely, didn’t I? I meant that companies like Yokozuna will most likely only deal with gaming companies and without entering into any exchange with the players themselves. They will probably ask the gaming companies as data controllers to provide sufficient assurance – including proof of consent if necessary – that they can process the data for the requested purpose.

plasmajohn
Reader
Patreon Donor
Loyal Patron
plasmajohn

Manipulative sales tactics, whales only pricing, and optimizing short term profits are band aids and more likely to cause churn than solve it. Players are getting wise to these and it erodes trust.

You want retention, make a good game, engage with your customers, treat them like you value their business. Forge long term relationships.

While trying to design by “gut feeling” isn’t the best idea, blindly following metrics is equally bad. I’ve seen some horrible design failures due to following metrics without proper root cause analysis.

Reader
Bryan Correll

Whew. I was worried for a minute there that the Yakuza was taking over the Japanese gaming industry.
#gaijingetswordsconfused

deekay_plus
Reader
Patreon Donor
deekay_plus

now this is the kind of thing i wanna see from GDC! thank you! :D

deekay_plus
Reader
Patreon Donor
deekay_plus

anyways machine/deep learning is super neat, and the limits of a project are really how much hardware you can afford to throw at it.

which is why sundial shoots down most of my project ideas pretty quickly as he’s only got 2 titan blacks to throw at our deep learning things we play with. tho we still end up having fun with the field, which will help sundial in his future career ambitions hopefully.