GamesBeat: It seemed like during this time, mining data meant real money. You’re saving money or you’re finding the money. Microsoft almost seemed like you might have been in the geekiest of jobs that nobody cared about, but at Facebook, there’s this appreciation for data science or an understanding that it can lead to money.
Berry: I know it’s going to sound weird from the outside, but when I was at Microsoft, when we’d sit down and talk about features, pretty much the first question was, “How do we monetize this?” At Facebook, it was never about that. It was always about, “How does this add customer value? How does this make the world more open and connected? How does this make the site better?” It was never about how to make money. If you get to scale, and Zuckerberg used to talk about this — when we get to a billion people using the site, we’ll worry about the money. If you build it, they will come.
When you get to scale, there are obvious ways you can make money. But we’re not going to design the features to make money. We’re going to make the site utilitarian. They used to talk about Facebook as a chair, or a light switch. There’s a switch on the wall, you hit the switch and the lights go on. You don’t think about it every day. It’s a utility. But if it wasn’t there, you’d sure as hell miss it.
I don’t know about you, but the first thing I do in the morning is to check my Facebook feed, see what my friends and family are doing, get the news. It’s a utility. When it’s got to that point of being a utility, you’d miss if it wasn’t there. Then you find ways to monetize it.
GamesBeat: When did it feel to you like data science was being appreciated for what it could deliver? Was it at Microsoft, or was it more at Facebook? Where I started to come across it was with Facebook games and mobile games. People understanding those analytics in free-to-play games, in particular, were the key to figuring out just how to make money, how to turn a game that had some deep flaws into a game that would last for a long time.
Berry: When I left Microsoft, there was no title in the address book of “data scientist.” I joined Facebook and there was a title there. It’s human nature to walk the edge of the curve and make things better. Out of all things, data science and analytics can make a good game better. They can’t make a bad game into a good game, but they can make a good game into a better game.
I think things are even changing further still. I gave a talk last year. It’s players who are becoming payers, not payers who are becoming players. Once upon a time, people used to walk into shops and buy retail games. You put your $49.99 down to get Flight Simulator, and either you suffer buyer’s regret or you’re really happy. The money would go upfront. Then we went through the web-based gaming, Flash games with adverts on the side. When I was at Microsoft it was one cent per user per hour, if you amortized everything out. If you could get someone’s eyeball for an hour playing a game with a bad advert, you made about a cent. Then the deluxe downloads came along, try before you buy.
There are exceptions, but now the world has moved to the free-to-play game, where the game is an advert for itself. You play the game and after a while either a pay gate comes along or there’s an option to pay a bit more. If you’ve had a bad experience, you’re not going to do that. I was talking to Jesse about this earlier. If you put your $10 down to the latest Star Wars movie and you come out and you’ve had a good time, you’re not upset about spending $10. If you’re playing a game online and you put $10 down and you get $10 of entertainment, you should be happy as well.
If you came out of a Star Wars movie and someone said to you, “Hey, for another $10 now you can see the next episode now,” you might say yes. You get more content. Again, the gaming business is entertainment. If you play an online game and you’re having a good time, you might put $10 down and get more content. It’s a good thing. It’s players becoming payers.
It’s really important to get things right. Analytics and all the rest of this stuff is one of the ways you track and measure everything that you can possibly think of, to find out what it is that people want to do, what makes the game better. There is a subtle difference between data-driven and data-informed. You can be slightly narrowed if you’re a data-driven company, where you only make decisions based on an instrument. There’s a quote attributed to Ford, although I don’t think he said it. If you ask people what they wanted, before the motorcar, they’d say faster horses. If you only ever think about doing something instrumented, you miss out.
Data-informed is probably a better way of describing it. You measure things that tell you about trends, but it still requires some thinking outside a purely programmatic-driven thing. You can miss that creative spark. You need to make sure that it helps drive the decisions, but not necessarily dictate the decisions. That’s the most important part of the data stuff.
GamesBeat: What are some gems of data science that come to mind when you think about what convinced people that data really matters?
Berry: It’s not so related to gaming, but I remember the famous thing about shopping cart analysis. They were looking through people’s receipts. Based on the amount of beer and diapers and milk, you can predict all this information about who’s in the household, the number of children and all the rest. Somebody said, “Let’s predict who will buy ice cream.” They built this model and this neural network and fed all this test data in and narrowed it down to a score. Then they put live data in and got garbage results on the outside. It didn’t seem to make sense.
It turns out that at the end of the day, the secret is, people buy ice cream when it’s hot outside. They hadn’t put the weather into the network. They were measuring all this and fitting a curve to noise. Making sure that you’re capturing the right parameters in the first place in order to drive things — understanding what the goal is, I think is important as well. For all the machine learning style of things, you need to feed that loop. Here’s the parameter. This is what it is we’re trying to do.
Sometimes that’s not as obvious. The gaming world, it’s even more so. It goes through — what the goal? Is it to improve ARPPU, number of users, all the rest of this stuff? There are interesting network effects that are missing if you just chase one thing. In an online game, a good chunk of the people will never pay. Only a certain percentage of people pay. But the un-paying audience is such an important part of the ecosystem.
A pair of jeans is $10. A pair of designer jeans is $200. Why do people spend $200 on Britney Spears jeans when $10 Target jeans work? It’s fashion. If you’re on a desert island and a pair of jeans washes up, you don’t care where it comes from, but when you go out in the street at night, people say, “That’s a nice pair of jeans.” That makes you feel good. If everyone was blind we wouldn’t care about fashion. It’s the same in online games. If I want to buy a red hat in a game, or a flaming blue sword, it might not change the balance of the game, but of that audience wasn’t there, there wouldn’t be a justification for spending money on that.
Designing a game, you need to make sure that you design it for everyone who’s in this chunk of the market that creates the ecosystem, the people who want to spend. It’s a very strict power law. As you know, the people who spend money on games spend it in egregiously large amounts. You talk about the long tail, the thousands of people who spend 25 cents, but that’s nothing. It’s the tall tail, the people spending more — 99 percent of all the money spent on Facebook games are people who spent more than $10 individually.
GamesBeat: Do you remember a USC professor, Dmitri Williams?
Berry: Yes, I know Dmitri.
GamesBeat: He was talking about social whales. You would miss the whole mark or point if you only looked at people who paid as important. He identified this group of people who were not spending money, but who caused other people to spend money. They were the life of the party, or the person at the bar who gets everyone else to ante up.
Berry: Or patrons. They buy suits of armor for everyone in their clan.
GamesBeat: Identifying those social whales was part of why the non-payers were so important.
Berry: It’s horses for courses, and it also depends on the kind of game. If you look at candy-like games, the way you look at the spending patterns, people play and play, and then they get stuck. They reach in their pocket, get the magic quarter out, put it in, and get a power-up gem or super cube they need. They spend money, go past that level at some cost, and they never visit it again. If you look at the spending patterns, there are lots of transactions, but they’re very small. They tend to be roadblock removers.
In a farming-like game, you buy a chunk of money, burn through that cash, and at that point, you decide, “Am I getting value from this game?” Then you ante up again. You tend to find there are much higher-value transactions with a longer time between them.
Farming-like games, people tend to put a lot of effort into it. They’ve converted real cash into assets in the game. Once they’ve done that, they’re more likely to not let their crops wither. They’ll keep going back to it. They tend to be more loyal. One of the things I looked at early on at Facebook was the number of games that people would play in a week. You found that if people played a farming game, they were almost monogamous about the game. When they had free time they played that game to the exclusion of other games. Slots players would go to a thing, pull a lever, run out of energy, and churn to another one. People who played casino-like games had no loyalty.
In terms of spending, we talked about — there are vanity items in games that don’t change the balance in the game in any way, but they’re important. Again, depending on the game, 20-25 percent of the revenue can be from things that don’t change the balance. That all happens because it’s a social environment. If you play solitaire at home and nobody else can see you, would you pay money to change the backs of the cards? No, you don’t care. But if you’re in an environment where you can spend $5 and the backs of the cards are all Amex Platinum or whatever, that shows your personality to everyone in the room, you might think of spending it. Just because of moving into a social environment — the game doesn’t change if you wear blue armor or red armor.
Then there are the time accelerators and the idea of cutting down on the grind. Kill 10 orcs to get to the next level. You can kill 10 orcs, but you have to find them. My son has finished his homework, he can go online and grind through killing orcs. I may want to keep pace with him, but I’m busy at work, so I’ll pay a gold coin for every orc I don’t have to kill. It’s not that I can’t kill them. It’s just adjusting the balance of the game. Put the windmill down and it’ll be ready in two hours, but I want it now, so I put the money down to build the windmill.
It’s very important as well, in terms of the offers that you go through — Bill Gates walks into McDonald’s, he’ll spend $2.99 on a Happy Meal because that’s all there is to spend his money on. If he goes to a fancy French restaurant, he could spend $10,000 on a bottle of wine if it’s available. If it’s not available, it’s not there. But if you divide one by the other and work out the average, if Bill Gates happens to be in that restaurant that day he moves the average spend — you don’t design your menu for people spending an average of $1,000. But if it’s available, someone can come in and spend a large amount of money.
The third one is harder to control, which is monetizing revenge. For the next 20 minutes, I want a 10 percent attack bonus. If you overpower that, then rich people buy their way to the top, and that isn’t fun anymore.
GamesBeat: Was it Clash of Clans where that was the tagline?
Berry: AJ Glasser was the first person I heard say it. I don’t know where she got it from. She’d probably claim that. But the skill you can buy from that has to be less than the skill you’d normally have otherwise. It wouldn’t be fun anymore.
GamesBeat: What did you think of when all of the whales would band together and have their boycotts of different games? I’ve come across that so often — Zynga, Machine Zone, Scopely, Kabam. It feels like it’s when they’re starting to hit this curve of decline. They either have to accept that decline, or they start to put the squeeze on the remaining players and make it twice as expensive for them to do whatever they’re doing. It’s not science to me. It’s just my observation. But I started to see games where these people were the most loyal players, playing every day, playing many hours, having all their friends in the game, and then at some point, they would feel squeezed and get fed up and stage a revolt.
Berry: I don’t have any direct experience. This is more thinking on my feet. One would hope that people running the game know what it is they’re doing. It’s the total area under the curve that matters. There was a horrible time in the games business where the paywall came up straightaway. You played the first few moves and the money came along. To me, that was kind of like used car salespeople. Fool me once. You might make a bit of money now, but you won’t make any in the long term.
Again, players become payers. If you provide a good experience, quid pro quo. They don’t mind paying. But it’s the area under the curve. It’s shortsighted. If you price people out, then you miss out on this integral going further down. If they’re doing that, maybe they’ve decided they want — it costs money to keep servers online. Maybe they’re trying to sunset the game and milking it for what it’s worth. It seems very shortsighted for them to ramp the price up just to make short-term money if they’re going to hurt themselves in the long term.
GamesBeat: It feels like older games are where this happens.
Berry: Maybe it is. You’re going to shut the game down, so — the vast majority of the money does come from the whales. So yes, if you upset those people then you’re not getting anything.
GamesBeat: A lot of the time the companies would say, “Well, it’s just a very small part of our base that’s complaining.” They would hint that it wasn’t significant, either in terms of the number of players or the amount of money. But it always struck me in that they were the most loyal players. It’s like saying you don’t care about your VIP customers because they’re a small group.
Berry: All the data, as much as I’m allowed to show — all the money comes from hyper-spenders. Something like 90 percent of the money in online games comes from people who’ve individually spent more than $5,000 on that game. If you’re spending less than $5,000 lifetime in a game, you’re in the bottom 10 percent. It’s a sharp power curve. It’s into six digits, the amount of money some people have spent.
If you go to Vegas and you’re a high roller, sheikhs pushing million-dollar chips on the table, those people have Facebook accounts, and one in three plays games. If they can push a million-dollar chip on the roll of a die, spending $100,000 on a farming game is nothing to them, and they do it. That’s where a big chunk of the money comes from.