Hacker News Reader

Comment by xyzelement | Hacker News Reader

Blameless is often pointless because sometimes something about the person is the problem. If the project failed because Bob ran it and Bob is too risk averse, then you can't fix it without talking about Bob. Bob either needs to figure out how to be less risk averse (hard and time consuming journey) or Bob shouldn't run projects that require risk taking.

Doesn't mean Bob is bad or gets fired, but he is part of the picture.

The whole blameless thing is so weak - if there's something about you that caused the failure, don't you want to know?

View on HN

The key mental shift (for me anyway) is that if a system can be brought to its knees by a single person, then the system is very likely flawed. You need to design a better system when the flaw in the system is the people. What that often looks like is changing/instituting processes such that quantitative measure (metrics, checklists, etc) governs decisions (thereby removing much, but importantly not all, of the human element), or you design processes in such a way that one person is not in charge of making the decision (the "two person rule", CRs, leadership approval). There are of course other tools but these two are pretty common in my experience.

> if a system can be brought to its knees by a single person, then the system is very likely flawed.

That just means that the person who designed the system deserves the blame.

I'm only half-joking here. You can't just rely on the "system" – someone needs to be responsible, either for the decision or for creating the system that makes the decision.

Having a system/process is not about removing accountability, it's about reducing discretion/cognitive load where it's been identified as risky. In fact, having a system/process in place to point to and say "this individual did not follow the steps/process/rules" makes an unbiased conversation about their performance much more possible.

Yes and this works when you're doing something for the 10th time. It totally doesn't work when you're doing something innovative and risky, which I assume is the kinds of conversations we're talking about here (this subthread is contextual to a senior amazon exec, he's probably not PMing someone forgetting to change the backup tapes)

If your root cause analysis leads to a preventative fix that amounts to “humans should not make mistake X” you haven’t done anything to prevent recurrence.

I am not constraining my statement to the narrow set of problems where your statement applies.

Reply on news.ycombinator.com