Q&A: Game Publisher Scores a Win with BI Analytics
In the hyper-competitive world of massively multiplayer online games, quick access to data is helping gaming company En Masse Entertainment make the right decisions.
- By Linda L. Briggs
- September 4, 2012
Massively multiplayer online role-playing games (MMORGs) are video games in which a large number of players worldwide remotely interact with each other in a virtual game world, using roles they assume for their characters. The games are hugely popular -- World of Warcraft, a popular MMORG, reportedly has millions of subscribers.
Another example of such a game is Bluehole Studio’s TERA, which was released in North America earlier this year. En Masse Entertainment, a U.S.-based game publishing company, spearheaded TERA’s release in North America. Although En Masse is just two-and-a-half years old and employs fewer than 100 employees, its use of data and business intelligence to make quick alterations even as games are being played shows a cutting-edge understanding of the role data can play in managing products in real time.
Jon Tuitte is a BI analyst at En Masse; BI This Week spoke with him recently about how En Masse uses QlikTech’s QlikView to analyze its gaming data.
BI This Week: As a BI analyst at En Masse Entertainment, what do you do?
Jon Tuitte: I’m specifically tasked with taking data from the various disparate data sources within our company and bringing it all together into an intuitive interface that people can use to make decisions on a daily basis. We’ve been using QlikView [at En Masse] for about a year for that purpose; I’ve been using it for about two years now.
How did you choose that particular product?
We looked at several [products] to see which would fit best here. What I liked about QlikView was its in-memory analytics capabilities. That allowed users here at En Masse to be able to slice and dice the data in a very quick fashion to get the answers that they need to make decisions.
How real-time is the data that you’re dealing with? Do your users need to make decisions very quickly using it?
We typically have a 24-hour delay on the data. Some of the data is more real-time than other bits of data, so it’s a mix. But for the most part, it’s delayed about 24 hours.
You mentioned disparate data sources. What kinds of data are you pulling into QlikView and from what sources?
As a game publisher, we have a number of different systems that our users interact with. We have a website, a storefront, an account-management system, and the game itself. I go to each of the data sources for these systems and pull it all into one data warehouse. The final step is displaying that data in a usable fashion.
What’s the software behind your data warehouse?
We store everything in SQL Server.
When you first started using QlikView, how challenging was it to get it installed and running for users?
It was very easy. One of the things I really like about QlikView is that they solve a lot of problems with you -- without you even knowing that [the problems] exist. They have a very simple installer that you throw on the server and run. You don’t have to do any kind of weird configurations or anything like that. It just works. The same thing goes with the desktop application. You just run the installer from the website and you’re ready to go.
What about beefing up your infrastructure? Did you need to add any hardware?
No. Currently, we’re running it on just one beefy server. As we scale up, we can build that out to more servers that can handle more dashboards. With the amount of dashboards we have right now, we need just that one server. It has about 92 GB of memory; that handles all the users within the company using the software at the same time without any problem.
How many users are on QlikView at one time?
Typically, we have about 25 people, which is about 25 percent of the company.
Who are the users?
We’ve deployed QlikView out to all the various disciplines here. What we want is for people making daily business decisions to be informed. We have the producers, we have a security-and-abuse team, we have game designers. They have various job functions and are looking at different datasets. They all log in at the same place, the same dashboard, but for different reasons.
How many dashboards do you have so far?
Just have two -- one main dashboard that has pretty much everything on it, and one that is specifically for customer service data.
You and your team created those dashboards using QlikView, right?
Yes.
How hard was that?
It was very easy. QlikView definitely makes it simple. I’m self-trained [in using QlikView] -- I didn’t take any training classes. I learned everything from QlikTech’s customer-support portal, as well as just reading on various websites.
What kind of devices do users display your dashboards on -- are they using phones and tablets yet?
I haven’t [deployed] the mobile side yet at En Masse. I know it’s easy, but we first have to design the dashboards to look good on an iPhone. As of right now, most people access the dashboards directly through the AJAX client.
But the dashboards could be configured for mobile devices fairly easily?
Yes. Actually, you can log in to the dashboard from your iPhone now. I’ve had one user do it -- it just looks ugly.
Within QlikView, what kinds of data dimensions are users looking at? Can you give some examples of how people are using it?
We have producers that are looking at data to see how people are playing the game -- which features people are using and not using. They might then discuss, “People aren’t using this feature. ... Maybe we shouldn’t spend any more time on it,” or, “Wow! This new item in our store is selling like crazy. Maybe we should create more items like it.”
Another use case is our marketing team. They’re able to log into the dashboard and look at all our various marketing campaigns: how did people learn about and first enter TERA? We can see exactly which websites, magazines, and various campaign methods convert better and which ones we should put more money into.
Have you been able to use agile development processes in developing dashboards -- has it been an iterative, back-and-forth process with users?
Yes, exactly. Many times, I can meet a 24-hour turnaround for a user request. That’s because QlikView makes it so much easier, from a development standpoint, to go out, find the data people want, and display it in a useful manner. Just putting a bunch of data in a grid is not really useful for most people. They need to be able to visualize it [graphically], then understand what they’re looking at.
How are you using QlikView for fraud detection? I know that can be an issue with gaming companies.
We can’t go into too much detail because we just don’t want to give away our secret sauce, but we do use it very, very heavily for tracking down malicious players within our game and removing them so they don’t become a problem for our customers.
How large is your database?
Before my ETL system runs, it usually churns through about 2 TB worth of data, maybe, on any given day among all the different servers. Then it compresses that down. At the end of the day, my big dashboard with all the data on it is only about 5 GB. It stays pretty small.
How is the data compression done?
It’s a combination of selecting actionable data using the ETL process and then storing it, as well as some sort of aggregation on the data, so it’s not always sitting at its lowest level. Then, in addition to that, I use the QVD files in QlikView that take data from your database and put it into temporary flat files. [A QVD, or QlikView Data file, contains a table of data exported from QlikView. Compact and optimized for speed, it is a native file format, that can only be read from and written to by QlikView.] When the dashboard refreshes, it refreshes from these flat files, as opposed to making a connection to the database and pulling all the data back in, so it’s almost like two ETLs sitting on top of each other.
What plans to you have for the product in the future?
We’re going to continue to refine and work on the existing dashboard to get it to a place where we feel very comfortable with it. We’ll keep refining it until it provides us with everything that we need -- we really want to try and make the data more accessible to our business users.