Strong opinions, weakly held

Automating accountability

In business, people talk a lot about accountability. In essence, it means verifying that people do what they’re supposed to, and addressing it with them directly when they do not. There’s a big problem with accountability, though, and it’s that it’s not easy to hold people accountable. Telling someone that they’re not performing up to snuff is one of the most difficult jobs a manager has, in part because it’s uncomfortable and in part because it has to be done in such a way that it leads people to improve their performance. (Unless, of course, they’re being fired. That’s even worse.)

World of Warcraft blog runs an advice column for guild leaders, and last week’s question was from a guild leader who needs to address a performance problem with a member of his guild. The details are amusing, but what really interested me was the proposed solution — using an addon for the game called Failbot.

Here’s the description:

Add some public humiliation to your raid! FailBot reports in raid chat (or another channel of your choosing) whenever a raid member “fails.” These fails are only things that are preventable by the individual on a consistent basis, so if someone’s name pops up it was irrefutably their fault, which makes it a great way to see who in the raid needs to improve without having to personally call them out.

I’m fascinated by this tool. It makes me wonder how we might develop similar tools for software developers. Clearly automated regression testing suites are part of the answer. I’m also trying release tracking tools like Pivotal Tracker.

I wonder what the similar tools are in other fields.


  1. This brings up the metric issue though. The moment you build a system that measures fails, people just adapt their behavior narrowly to avoid this criteria. Now you have trained people the most important thing is to not get called out for failing, regardless on whether what they are doing is productive or not.

  2. That’s the fundamental difference between games and real life. In World of Warcraft, if a rogue does 2000 DPS and avoids any of the “fail” conditions, they did their job. The measurable conditions perfectly model the optimum performance for a player. In real life, things are much more complex. That’s why systems that rate programmers on things like lines of code written or number of bugx fixed are such a disaster.

  3. I joke about being a fan of requiring developers who break the build to wear a helmet for the day. They’re clearly a danger to themselves and others.

Leave a Reply

Your email address will not be published.


© 2019 rc3.org

Theme by Anders NorenUp ↑