|
 Features
Postmortem: Blue Fang's Zoo Tycoon 2: Marine Mania
 The Build is Going to be Done When?
Build times were relatively long while we were developing Zoo Tycoon 2. When we did the first expansion pack, Zoo Tycoon 2: Endangered Species, they got a bit longer. And by the time we were working on Marine, they were longer still until ultimately it took more than 4 ½ hours to complete all the builds that were needed for our milestone delivery. Although we’d looked in ways to improve the situation, a clear solution didn’t present itself. We mitigated the situation by becoming creative about when we kicked off builds and we did nightly builds religiously, but it was still a hassle, particularly during milestone weeks.
It was clear that no one on the team actually liked how long they took and it would be a regular topic of complaint and discussion. Yet in spite of investigating various options, a method to speed them up didn’t materialize. To compensate, we established a scheme to do only one of our required builds on milestone day, test it and only once it was confirmed “good” would we kick-off the other builds.
Early concept art
After discussing it with our Program Manager at Microsoft, we also had the OK to fall back on sending only the one build and uploading the remainder later over the weekend if necessary. Ironically, we never needed to actually use this scheme as we always had the full builds ready to go sometime on the Friday when they were due.
Close to the end of Marine, our IT manager took up the challenge. After he did some experiments with various hardware options we ultimately cracked open the piggy bank to buy some tremendously serious hardware, and migrated our build scripts to the new environment. Voila! About three weeks out from our Zero Bug date, our IT manager successfully solved our build time problem and our build times dropped to under two hours. This was incredibly helpful as we kicked off a lot of builds between then and when the product was released for manufacturing. And it goes without saying that he was quite the “hero on campus” for quite a while after that!
Can Everyone Hear Me?
We introduced a concept that we called “Feature Groups” in Marine, and really the idea was to simply formalize the way people tended to work anyway… that is, by forming small groups of people from art, engineering and design who worked closely together to implement features. The idea was that feature design, task analysis, implementation quality review, etc. would all be performed by this group. But the groups ending up including anyone who might ever work on that feature, so they ended up being quite large.
These large meetings often felt inefficient and poorly focused. Instead of zeroing in on the tasks at hand, they’d often turn into brainstorming sessions, and one thing we didn’t have at that point was a lack of ideas!
We did end up reining them in, and it took everyone’s effort to keep them on an agenda and stop them from going down the inevitable tangents that would come up. In the end, we fell back onto the way we always tended to work and the groups went back to being smaller groups comprised of the key people working on the feature.
In a way though it wasn’t all bad because we did draw some good ideas from the unexpected brainstorming, and the team members who participated ultimately had a much clearer idea of the product we were building.
|
|
|
|
|
Previous