Every automation has two costs. The one everyone counts is building it. The one that kills it is maintenance, and maintenance is proportional to how many things outside your control it depends on.
A script that reads a file I control and writes a file I control has run for years without attention. A script that reads three APIs, two of which change their response shape without warning, needs someone every quarter. Same effort to build, completely different lifetime cost.
So the question I ask before building is not whether it saves time. It is how many things outside my control it touches, and whether anyone other than me can fix it when one of them moves.
The place I get this wrong is client reporting, where the pull is strongest and the dependencies are worst. I keep rebuilding the same fragile thing and I keep being surprised when it breaks.
What have you automated that is still running untouched after a year, and what made that one survive?
Swiss Knife SEO ModeratorsModeration team