Q&A: Zynga Stresses Visual Access to Its Vast Data Volumes
Online Social Gaming Company Turns to Visual BI to Delve into Data
- By Linda L. Briggs
- July 21, 2010
With over 230 million users monthly playing its games -- including FarmVille and Mafia Wars -- Zynga is one of the largest online social gaming developers in the world. In this interview, Ken Rudin, general manager of analytics for Zynga, discusses the challenges Zynga faced in getting meaningful insights to its analysts on several terabytes of game data captured daily -- and the data visualization solution the company recently selected.
"I've been in the large data world for a long, long time," Rudin says about his decision to join Zynga last year. "I absolutely love the technical challenge when you get into massive skyscraper-size chunks of data. Plus, I love even more the business value you can get out of having that much data… If you know what questions to ask, you can get some phenomenal answers."
BI This Week: What led to your decision to join Zynga last year?
Ken Rudin: When I was introduced to Zynga through a friend, his introduction was "First, these guys make games so you know it's going to be fun to work there. Second, they have one of the largest data warehouses in the world. It's growing incredibly rapidly and the entire company is essentially run off analytics and insights." That sounded like a perfect fit for me.
What he told me has turned out to be true -- we have incredibly high usage of our data throughout the company. On any given day, a very high percentage of our employees are looking at the data and running analytics. Our analysts have a strategic view of how the company works. Our managers ask for real insights daily. Instead of being asked "Can you run this report for me?," the questions are more along the lines of "What should we be doing about our e-mail distribution?" or "Can you take a look at that and give me some recommendations?"
You're running one of the largest data warehouses in the world, aren't you?
Yes. Vertica is our back-end solution; we are the largest Vertica implementation in the world.
What led to the decision to go with Tableau?
I'm a strong believer that the real value of analytics comes from two things. You have to be great at the science of it and you have to be a great communicator. Where analytics fails in most companies is not the science, it's the communication of the results. People come out with tons of data but they don't know how to describe what it really means to the company and, more important, what they should do about it.
I love getting asked the question here, "Can you look into this and come up with a recommendation?" versus, "Can you make this report for me?" The important thing is the data as well as the communication of what it means.
Before Tableau, there were a lot of tables and numbers, but it's very hard to communicate those results. You always needed someone with you to describe what the numbers on the table mean, or to write up long discussions of it.
We were looking for something that would address two needs. First, give us a visual way to communicate. Particularly with numbers, people are much better at understanding data when it's presented visually versus as letters and numbers. We wanted something that was great at that.
Second, we wanted something with interactive capabilities so we could look at the data and say, "Why is there a bump there? Why is there a dip there?" and then drill into it and come up with an "Aha, that's why" moment. We didn't have that capability before.
We had a Vertica data warehouse, and you could write SQL statements to go against it and get your results back in a table. The answers were there, but again, there was no good way of communicating it easily.
We use Tableau a lot internally. It's one of our primary tools right now; it's as much an analytic tool as it is a communication tool.
Who uses Tableau at Zynga?
There are two types of users here -- I'll refer to them as the producers and the consumers. The producers actually create the dashboards and the analyses -- there are a couple of dozen people right now doing that.
The consumers -- that's a very large number. A very high percentage of our employees are consuming the information via dashboards and portals every day.
That's always an elusive goal with BI, of course -- to get as many users at as many levels of the company involved as possible.
What we're doing works great. People can look at a chart and it's fairly obvious what's going on. One of Tableau's strengths behind the scenes is partially automating the best way to visualize a particular set of data. It's one of their secret sauces. You don't want a bar chart for everything.
Tableau makes it easy to uncover insights, and then picks the best way to present them visually so they can be easily understood. We can now put the data out there and I don't have to go and talk to 250 people individually to explain it to them. An intuitive chart does it for me.
Did Tableau integrate with Vertica fairly seamlessly?
Yes, we didn't have any issues there. That was one of the things we checked during trials. Vertica also recommends them, so it was a good fit for us.
When you looked at other products, what in particular made Tableau a good fit?
First of all, as I said, it did the best job of helping with visualizations. Also, many of the other tools we looked at required you to either work with just a subset of the data or pull a full copy of the data into its own local storage. With the first approach, you don't get all the data, and due to our data size, the second approach wasn't feasible. Tableau is one of the few that goes back to the source, the main database itself, when you issue a query, versus having to pull over a cube.
We looked at traditional BI players, and you have to pull information into some cubes to get your work done. We saw one [product] that you could query directly against the database with, but the tool was just too complex for our users.
I guess that's a plus and a minus. Even though some companies have extremely powerful analysis tools, we really wanted an easy-to-use tool for communicating insights. Tableau has all the analysis we need. … I'd rather have good analytics and great communication than great analytics and poor communication because in the latter case, you get nothing out of it. I don't mean to imply that Tableau is at all weak on the analytics part of it. They do everything we need and more.
You've stressed that communication aspect several times. Have you conducted training at Zynga on using Tableau?
[We haven't had to train much at all on using Tableau, but] one thing that has been very helpful for us is training our analysts in visual communication. There's written communication, and there's visual communication -- how you get the data to tell a story. Tableau is a great tool that allows you to tell great stories, but you can tell either a good story or a great story -- a lot of that comes down to the skill of the storyteller.
Our analysts need to be technically smart and also great storytellers. When you put out a result, it has to say: Here's what we looked at. Here's why we dug in a little deeper because there are three things here. We excluded these two which left us with this one. Then, we drilled into that a little further and here's the nugget we found. You just walk somebody through the process so they understand what you looked at, they understand the logic behind it, and they get the result.
We've done some training internally on how to communicate visually and we're also bringing in one of the godfathers of visual communication, Stephen Few, to do more training.
That's a great idea.
Again, we're trying to push the limits of communication of data, not just generation of data. So many tools focus on how you generate the data, but we'll spend 50 percent of our time generating the data, and the other 50 percent focusing on how you communicate it so that people act on it.
That's what Tableau's strength is. It makes it much easier to get people to act. My fundamental philosophy is, if you're doing analysis, you might be generating fantastic insights, but if you distribute your results to people and they aren't able to understand it, they won't actually do anything differently. In that case, by my definition, the analysis has added zero value to the company. You can say, "We did all these great analyses and there are all these things we uncovered," but the bottom line is, if the company does nothing different, it's hard to argue that you're adding any value. To get somebody to do something, [it's often in] the way you communicate…
Using Tableau, we show people what needs to change and what the outcome is going to be. It's all there in the charts and graphs.
Tableau refers to itself as a visualization tool. I take the communication part of that. It's not just visualization in order for our analysts to see what's going on -- it's the ability to communicate those results to consumers so they a) understand it, and b) will act on it.
Can you give me an example of how you're specifically using Tableau at Zynga?
We launch new games all the time; we want to look at the impact of each new game on our business overall.
Let's say you're playing FarmVille right now and we introduce Game X. It quickly gets a lot of players, and we want to look at where those players came from. How many of them were new to Zynga? How many of them play FarmVille and now also play Game X? How many of them used to play FarmVille, have stopped playing and now only play Game X -- meaning we cannibalized our own user base? How does that trend over time? How does that compare to other games we've introduced?
We also want to slice and dice player demographics. How many of them are playing every day? Three to five times a week? Once a week only? How is that changing over time?
There's also the viral aspect. … Not only do people play the new game but if they love the game they send out messages to their friends, saying, "Hey, play that game with me because we can be neighbors in the game and then we can help each other achieve goals within the game as a team." We measure that. We want to see what kind of people came in early and who were the ones most responsible for the growth of that game. …
That's the whole idea. We're not just looking at games in isolation, but rather we're looking at a broad perspective of how a new game impacts all aspects of Zynga.