"is it worth the time?"

ops cultureprogramming

The question of whether automating a task actually saves more time than it costs to automate.

A famous xkcd table weighs how often you do a task against how long automating it takes, to show when the effort pays off. It is the honest counterweight to the engineer's instinct to automate everything, and a close cousin of yak shaving.

Randall Munroe's Workflow comic shows a user complaining that a keyboard shortcut has stopped working, and the maintainer discovering that the behaviour was an accident being relied upon by someone whose entire process depends on it. It is the most compact illustration of Hyrum's law available.

The comic works because both parties are reasonable. The maintainer is fixing something genuinely broken; the user built a workflow on observed behaviour, which is the only thing users can actually observe. Neither consulted a specification, and the specification would not have helped, because it did not describe the behaviour either way.

Its practical use is as a shared reference during API design arguments. It reframes the question from whether the old behaviour was correct, which is usually not the interesting part, to who depends on it and what happens to them, which is. It also implies the mitigation: if you cannot know who depends on an accident, you can at least stop creating new accidents, by making unspecified behaviour visibly unreliable rather than quietly consistent.

Also known as: automation tradeoff, xkcd 1205, worth the time

← All glossary entries