Blizzard gives Battle.net post.mortem
GDC Online 2010: Project director Greg Canessa and technical director Matthew Versluys explain exactly why launching a new online service to handle the millions of World of Warcraft and Starcraft II players from day one is no easy task.
Who was there: Blizzard's Greg Canessa, project director for Battle.net, and Matthew Versluys, technical director for the service.
What they talked about: Battle.net was blamed as the reason for Starcraft II's delay out of 2009. With the game and the service now up and running for months, Canessa came to GDC Online to go in depth on the development of the online service from conception to launch.
Canessa started by recapping the backstory of Battle.net. It originally launched in 1996 with Diablo and was billed as the world's first integrated matchmaking service. It changed over time with the releases of Starcraft, Diablo II, and Warcraft III. Although it originally ran on a single box, with Versluys deadpanning that it runs on "considerably more than one box" today. As of a year and a half ago, Battle.net had 12 million players on it, with about half of the audience hailing from Korea.
A lot has happened since the service launched, with Canessa talking about how Steam, iTunes, PlayStation Network, and Xbox Live have revolutionized digital distribution, while Facebook and MySpace have changed the way people interact with one another. Things have changed at Blizzard as well, with World of Warcraft's success having a profound impact on the company.
For the new Battle.net, Canessa said the team had a vision to create the world's premiere online game service that would connect the company's entire catalog, with a goal to make Blizzard games (and their multiplayer experiences) more fun in the process. To achieve that, Versluys said the first step was to staff up the team. They built the service from the ground up and experienced a significant problem finding people with the proper qualifications to achieve their goals.
The first of the five main topics Canessa wanted to address was the idea of taking a deeply integrated approach to Battle.net. Blizzard wanted to create a seamless integration between the games and the game service. If players of Starcraft II don't know exactly where Battle.net stops and the game begins, then the team has done its job well.
One of the biggest lessons Blizzard learned was that it's "bloody expensive" to take this approach, requiring a complete rewrite of the client program each time a new game was tied in to it. Another lesson was that it can be challenging to work with the game-specific teams while the service is being created. Canessa acknowledged that the Starcraft II delay caused a good bit of friction between the game's development team and the Battle.net crew.
Canessa said he also learned how much work was involved with integrating social gaming features not just with the Battle.net framework, but also with the game. Making sure that things were integrated in ways that made senses (for instance, not popping up achievement notifications if someone is playing in a tournament) took lots of iteration and lots of extra work.
Next Canessa showed mock-ups of the new Battle.net's user interface at various points in development. In late 2007, there was an IRC-style chat window on the left of the screen, with the gamer card in the upper left. Throughout 2008, more browser-style features were incorporated, the gamer card moved to the top right, and a social bar appeared at the bottom of the screen. By mid-2009, Battle.net began looking more and more like an integrated Starcraft II product, with the main navigation in the top left. By 2010, the interface was largely recognizable as the current one, with tweaks to the color scheme and button appearances gradually phasing in until the launch.
The next big lesson the team learned had to do with the user outrage about a decision to make the Real ID system require a person's real name. For Real ID, Blizzard found there were issues with having multiple tiers of virtual identities--specifically, one name for each game and another for the Battle.net system. At the same time, the developers noticed that social media sites were using their real names. Given the security concerns with using an e-mail address as the Battle.net name (many users have one password for all of their accounts) and a desire to make players more accountable for their actions, they originally decided to require that real names be used.
That decision did not go over well with many gamers, and the requirement was changed. Despite that, Versluys said the ultimate community reaction to Real ID was more positive than they had anticipated. In the end, they learned a few things from the incident. They realized that anonymity is very important to gamers and that having tiered identities requires more work but was ultimately the right call.
And while players value anonymity, Blizzard realized that they generally want to make just their own information private but still take advantage of the features that rely on others' personal information. Without a way to see who a player's friends are friends with, it would have been too difficult and time-consuming for players to fill out their own friends lists. It's much easier to just look through their lists for common acquaintances and then fill in gaps by inserting other friends' e-mail addresses one at a time.
The third big topic of the presentation was World of Warcraft integration. Canessa said the goal was to introduce new features that WOW players could enjoy even if they didn't care about Starcraft II, things like cross-realm chat and cross-faction chat. Naturally, they learned that integrating with an existing community of 12 million users and not mucking things up is difficult. Integrating friends lists and different styles of chat were just two problems facing Battle.net, and the team quickly learned not to assume that the social features they were trying to implement weren't necessarily right for all users and all situations.
For example, Canessa said there was a huge customer perception problem on Blizzard's forums when they talked about Facebook integration. In reality, all they released was a friend finder app using the social network, but there was "a lot of sensitivity around social integration" in the user base.
Launch complexity was the fourth main topic of the postmortem. Versluys said the Battle.net launch was easily the most complex since he'd joined Blizzard a decade earlier. Blizzard launched the service time zone by time zone around the world in a single day, across five continents and in 13 languages, with millions of copies of the games going out through physical and digital distribution. For Canessa and Versluys, bringing it all together was like landing a 747 on an aircraft carrier. However, the game did launch, and in the process it lived up to one of Blizzard's core values as a company: think globally.
Canessa noted that the last time Blizzard shipped a retail box product without a recurring revenue stream was back in 2003, with the Frozen Throne expansion for Warcraft III. After seven years, it was an adjustment to get beyond the thinking of shipping an MMOG game and get back to the problems unique to the shipping retail titles.
The last major topic they touched on was building and growing the Battle.net team. In the last 18 months, the Battle.net team has quadrupled to close to 50 people, and there are still another dozen positions to fill. The team has also been reorganized three times in that span as their goals changed.
Canessa said it's tough to find new people because it requires a very specialized type of talent, and there aren't that many game services out there to draw from. Instead, Canessa said Blizzard had to take people who were strong from a core development standpoint and had some related experience--smart people who were passionate about what a game service could provide.
The bar for an online service has gone way up in the last decade, Canessa said. A lot of what people expect and demand from a game service, as well as microtransactions and reward systems, are all drastically different from 10 years ago. Additionally, launching the service is just the beginning. The team at Blizzard has plenty of new features planned for Starcraft II and is working on feature-set roadmaps for World of Warcraft and Diablo III.
Quote: "We have literally 10 years' worth of features on a roadmap for what we would like to do with Battle.net."--Canessa
Takeaway: At the risk of understating the issue, it's not an easy thing to create a new online infrastructure for a hugely anticipated game with a worldwide simultaneous launch and seamlessly integrate an existing community of more than 12 million people into that service, all the while finding people capable of making that happen.