Sunday, March 31, 2013

Second-System Syndrome

Close to 40 years ago, Fred Brooks wrote one of the classic tomes about managing computer projects, The Mythical Man Month. His best-known observation was that communications cost in a project rises exponentially with the number of people, so that adding additional people to a project doesn't speed it up as much as one might expect, and that beyond a certain point, adding people will actually slow a project down. This is best summed up in his pithy saying, "Adding more people to a late project makes it later." It's the reason that startups can outperform big companies, and it's as true today as it was in the 1970s.

But Brooks had other telling observations, and one of my favorites is about "Second System Effect." Brooks noted that teams which had been successful with their first project often failed spectacularly at the second. He hypothesized that during the first project, a teams would proceed cautiously and put aside most suggestions for frills and enhancements. Then during the second project, all these stored-up geegaws got piled into the design, resulting in, as he put it, "a big pile."

I'm seeing this at work in an entirely different arena, namely anime fansubbing. As I've noted before, the FroZen-EviL team that did Yawara defied the odds (and the lackadaisical tendencies of all three parent groups) by finishing 66 episodes in little over a year. The team did this by keeping the staff constant and small, and by maintaining a ruthlessly simple workflow.

FroZen-EviL was supposed to be a one-and-done for Yawara, but the experience was so enjoyable that almost everyone wanted to try another project. After some debate, the team chose Miyuki, a slice-of-life show from a decade ago that had been dropped by several different groups. Yawara's translator had to bow out for real-life responsibilities, but the team found another translator, who ripped through all 37 episodes in under a month. Further, the original Miyuki Fansubs scripts were available as a reference for the first eight episodes. Everything seemed good to go... and then, dreaded second-system syndrome appeared.

It turned out that everyone on the team (except me) had been harboring deep thoughts about improvements in the process and the outcome. So many things in Yawara could have been done better: real karaokes instead of line-timing; real typesetting instead of "{\an8}Sign:"; more QCs to get every last nuance correct and error out. So more people were brought to implement all these improvements. The results were just what you might expect. Whereas a typical Yawara script went through three revisions (original edit, QC applied, RC applied) and required less than a week from translation to release, the first Miyuki script went through at least twenty revisions and required close to two months. A textbook example of Second System Effect.

I think everyone is a bit chastened by what happened on the first episode. I believe that the workflow for subsequent episodes will be more disciplined. After all, the purpose of life is to develop good judgment; and good judgment is a result of bad experiences. At least I hope so.

1 comment:

  1. I never imagined it took twenty revisions to sub the first episode of Miyuki. I guess you really enjoyed the process :)