I once sat in a postmortem where the entire discussion collapsed into a single sentence: "Marek pushed the change without checking." Everyone nodded, the meeting ended early, and we all felt vaguely satisfied that the mystery was solved. Six weeks later someone else, following the exact same deploy process, took the exact same system down in the exact same way. Marek had left the company by then, so we could not even blame him a second time. The real cause had been sitting there the whole while, patiently waiting for its next volunteer.
That was the day I understood that a postmortem which ends with a name has not actually finished. It has just found somewhere comfortable to stop. Blaming a person feels like an answer because it closes the emotional loop. Someone was at fault, the fault has a face, and the rest of us are implicitly safe because we would never have been so careless. It is a story that protects everyone in the room except the next person who inherits the same trap.
The uncomfortable truth is that people do not cause outages on their own. Systems let them. If one engineer running one command at the wrong moment can break production, the interesting question is never why that engineer did it. It is why the system permitted a single tired human to cause that much damage with a routine action. The person is the trigger. The missing guardrail, the silent failure mode, the deploy step with no confirmation, that is the cause.
So I train myself to keep asking "why" past the point where it feels rude. Not to interrogate the individual, but to walk backward through the conditions until I reach something we can actually change. Why did the command have no safety check? Why did nothing warn before the damage? Why did we trust memory instead of a gate? Each answer points at the system, which is the only thing we can fix, because the next person will be just as human as the last one.
Name the failure, never the person. The moment a review turns into a search for someone to blame, everyone quietly learns to hide their mistakes instead of surfacing them, and you have traded one honest engineer for a whole team that stops telling you the truth.
– Serguey Shinder
Top comments (0)