3 Years of Managing the Guns of Icarus Community

3 years ago, when we released this online-teamwork-multiplayer-Steampunk-airship-combat game called Guns of Icarus, we didn’t think we would still have the privilege of updating and support the game today.  We released in the middle of a hurricane, with servers failing and us having just about no clue how to manage an online game community.  The game was near the top of the best seller’s list on Steam, and many people could not even get through Steam authentication (Steam was having server issues as well because of the hurricane) let alone connecting to our servers, which were also knocked out by the storm.  We didn’t know if we would make it through the first week.  

Over the last 3 years, we had our servers crushed by the mights of TotalBiscuit and other youtubers, and we’ve been ddos’ed.  In one stretch I slept so little that, in semi consciousness, I ran into a door beam and got a concussion and a nasty cut, spewing nonsense in meetings that no one understood.  Through it all, we’ve learned a few things that hopefully would be of use.  

One of the first things we learned from a rough launch is that, players look to the devs for issues beyond those directly related to the game.  Whether it was someone’s LAN setup, firewall, antivirus, VPN, malware, ISP, or issues related to Unity or Steam, players came to us because they didn’t have anywhere else to go.  By in large, even with early testing, player interact with the game near, at, or post release.  As developers, we had the mindset of looking at the release milestone as almost an end point.  As it turned out, our expected duration of interaction with our game was grossly mismatched with the duration of players’ interaction.  That was the first major mindset change we had to undergo as a team.  We realized that post-release was arguably more important than pre-release and release.  

Once players were in game, we were no longer just developers.  We were service reps as well, whether we liked it or not.  Once we realized that, there were two things we needed to do, and quickly.  First was to define for ourselves what it meant for us to serve - how do we serve, to what extent do we serve, and what principles do we live by.  The other was building a process as well as a support library and knowledge base.  

Service Model

With a paid download game, a player pays once, and can play for as long as he/she would like.  With an indie game on Steam, the price point generally ranges between $10-25.  Typically, the higher the price paid, the more premium the service.  For example, if I pay to ship a package with FedEx, I would expect better service than USPS (less wait, more reliable delivery, better forms, smoother and clean pens, nicer looking packages, etc.).  So the challenge then became, how would we offer good service while players weren’t expected to pay as much?  

We started looking at businesses in different service industries that would hopefully guide us.  Then, this dumpling chain in Taiwan called Ding Tai Fung (鼎泰豐) somehow popped into my head.  This place used to be this mom and pop place at the entrance of an alley, the quintessential neighborhood restaurant.  Over the years, they expanded and expanded, centralized their operations, systematized their service, and exploded into an international chain.  Price wise, it’s Chipotle plus, very affordable, but a bit more expensive than fast food or street food.  Food wise, it’s incredible mouth-watering deliciousness that rivals any fine dining restaurant.  It is in fact the only restaurant in Taiwan to have received a Michelin Star.  

The line at the restaurant is usually really long, but ordering is mostly done in line.  By the time people sit, food usually arrive within minutes.  Any request, from highchair to extra utensils, can be fulfilled by anyone, from busboy to manager.  

In studying their restaurant management and service, we realized one important thing - they are only governed by a few principles, and are not hampered and burdened by rules.  They tended to say yes a lot and rarely said no.  They were extremely flexible, personable, and fast.  No one hid behind rules, and everyone seemed empowered to act.  In short, they were principle driven instead of rules driven.  If we were to deduce the essence of their service to a few principles, we thought they would be speed, niceness, and empowerment.  The next question then became how that applied to us.  

We decided that speed for us meant turnaround time of player reports or feedback.  We tried to reply to everything our players threw our way within 36hrs as an ambitious goal, and we were able to by in large keep to that over time.  We were also determined to read through and process every player report in terms of moderation.  With niceness, that became a requirement regardless of how difficult or frustrating.  With empowerment, a couple of us were trying to do it all, and that became increasing difficult as volume of player feedback grew.  We therefore had to look for a way to scale the help and empower those who helped us to both moderate and reward.  For this we partnered with our players.  

Social Systems

In order to achieve speed, we needed a system and a process.  The design and systems of the game itself had to support our community management and service principles.  Guns of Icarus is a game where teamwork is required.  The communications system therefore had to be at the core.  In addition to hierarchical voice chat channels that facilitate crew-captain, and captain-captain communications, we also added voice commands for international speakers.  Our server structure was one where anyone in the world could play in the same match.  We have a global chat where everyone can see and talk to each other in game, and this turned out to be extremely effective for us to support players at the point of the issue as well as showing new players the activity of the community.  

Beyond core systems and features designed to foster teamwork, we looked to design and implement features focused on improving positive social interactions.  From a simple one time thumb up for crew, captain, and opponents at the end of a game to incorporating new player teaching and commendations into clan progressions, we tried to model and reinforce positive behavior.  For an entire clan to progress, members needed to crew with new players, to teach, and to give novice player commendations.  This created tangible and meaningful value of novice players to veterans, thereby creating a social based leveling and reward system.  

Another system we created was called Applaud.  I went to a high school that was supposed to be a diverse community.  It was diverse, but it wasn’t much of a community.  People isolated into cliques, and there wasn’t much to bring people together.  I look at my daughter’s elementary school.  It is equally if not more socio-economically diverse, and it is one harmonious community.  The difference I realized was in the little things made with conscious, deliberate effort.  For example, every week in my daughter’s grade school there is a chapel.  The chapel celebrates all religions and cultures.  During chapel, kids would volunteer to go up on stage and announce a good deed by a classmate or a friend that affects them in a positive way.  Those little gestures matter.  Taking inspiration, we designed a system where a player could nominate another player for sportsmanship and citizenship and tell us about it.  Every week, we selected “applauds” to post onto our forum.  Both the “applauder” and the “applauded” would be rewarded with gifts of game items from us.  

For us, positive reinforcement in this case was effective.  In the history of the game, there had been about 60K player reports.  In the first 3 months of implementing the Applaud system, we had over 30K applauds.  

Kick, whether captain or vote, is one of the most often player requested features for us.  Yet, we never implemented it.  In the game, tension could arise and tempers could flare. In an intensely competitive PvP match that requires teamwork, winning can be exhilarating and losing can be really deflating.  In looking at a significant number of player reports over time, we learned that, the definition of trolling for instance could be really gray, often depending on the perspectives of the players.  A pilot can accuse a crew of not being on gun while the crew was trying to make a last ditch repair.  These cases were far from black and white.  Had we implemented a kick function, it would not have helped in conflict resolution and would likely have led to more “trolling.”  Often friends crewed together.  Had we implemented votekick, the odd person on a ship would have had little to no protection in case of a dispute.  While we moderated obvious cases of hate, these types of player disputes at the margins of winning and losing were much tougher to moderate, and kick would most likely made the situations worse.  

Support Methods

Social oriented systems in game allow us to encourage positive social interaction and deliver help at the point where issues first arise.  To deliver speed in service, we needed a way to effectively manage player feedback and feed it into our production pipeline as well.  In the beginning, we tried to reply to player feedback and issues wherever they were raised, and our points of contact extended from in game, social media, and forums to Steam chat, Skype, and Kickstarter.  We tried out a QA reporting site as well.  Ended up, it became hard for us to keep track of everything and getting back to everyone in a timely fashion.  So we decided to simplify.  

We had players email their issues to one support email.  From there, a couple of us took on first responder duties, solving as many issues as we could, with the support library that we’ve been accumulating.  For issues we needed help with, we then pulled in the rest of the team.  Each time we solved an issue we’ve never seen before, we’ll add it to a shared doc and publish it to our FAQ.  We’ve spent an entire afternoon with a player to discover a particular router model corrupting UDP packets for instance, and that discovery ended up helping quite a few players with the same issues.  For player ideas, we logged them together with our own.  

We tracked the status of each player’s issue, and for bugs and features worthy of implementation, we specified them further and fed them to our sprint planning.  

We reviewed player ideas every sprint, and once we decided to pursue them, there was no longer a difference between player ideas vs. our ideas.  A good idea was a good idea.  

Over the last 3 years, we replied to and reviewed over 30K Steam forum posts, over 100K posts on our forum, and 51K support emails (totalling 24GB).  And we implemented over 1,200 features and fixes based on player feedback.  Initially, managing this volume took significant amount of time and effort for us, fracturing development.  Early on, we were fortunate to have quite a few passionate players who wanted to help beyond merely playing the game.  We decided to formalize player volunteering into a program called Community Ambassadors.  The program had a service and teaching mission.  We formalized an application for the players to describe their experiences in game, experience levels, and past experiences moderating or managing game communities.  We selected people based on citizenship and other players’ recommendations.  Once someone was a part of the program, the person was empowered to act and, to a large extent, became a part of the team.  We allowed them to be in novice matches to teach, gave them moderation functions in game and on forums, where they moderated different boards.  We opened our item generation system so they could reward people in game.

We wanted to ensure that they succeeded, so we held regular training with them, going over different support situations and cases.  We also shared our support library with them.  We set up group chat with them as well so we were constantly in communications.  They also had distinct status, badges, and title in game, so a new player could easily identify and find them if help was needed.  We also rewarded them for helping us, and they got news and updates about what we do first hand.  We’ve had show booths completely manned by player volunteers, and they host most of our competitive tournaments.  Over time, the program grew to over a hundred members, and over 50 are active at a given time.  

Service Mindset

There were plenty of moments in the last 3 years when our mindset and resolve to be nice in the face of hate were challenged to the core.  I have all the respect in the world for everyone in the service industries who tries to do a good job.  To serve is a noble profession, and to do it well is not easy at all.  I think a key to success is to treat it as a serious job, and divorce it from being personal.  Someone who says mean things and has unreasonable demands might just be having a bad day.  Our responses had the power to perpetuate the downward spiral, or to lift someone up.  The above quote was one of the worst emails we’ve had (and there were quite a few like that).  That this player was reported in game multiple times by multiple people was no surprise.  We then gave him a warning, which he confused to be a temp ban.  He then got incensed and wrote us an email.  

Initially, seeing all the text that I highlighted in red was tough to take.  I had to walk away for a bit to collect myself.  When I returned to the email and calmed down, it became easier to see the text that I highlighted in blue, which was the point he was trying to make.  He paid for the game and spent his hard earned money.  He liked the game enough to gift 3 copies to his friends.  He misunderstood the warning for a ban.  Once I understood the issue, it became easy to reply.  I also appealed to him to help us maintain and grow this community while reminding him of ours as well as Steam’s terms of use and community standards.  The player never expected me to reply, and once I did, he was nice enough to help us and turned his behavior in game completely around.  We’ve had quite a few players with a similar encounter, and the majority of these players actually stayed with the game over hundreds of hours of play.  With over 60K player reports investigated, we ended up giving out warnings only about 4% of the time.  Only 0.3% of the time did we issue bans of varying degrees.  We take players’ hard earned money seriously, and we try to appeal to people’s humanity and reason over heavy handed policing, and I think it has worked for us.