Most systems are designed from the happy path outward. The team builds the thing that works, then patches in error handling as an afterthought — a generic message, a cryptic code, sometimes nothing at all.
That’s backwards.
When something breaks, the user is already in trouble. That’s the moment the system owes them the most — clarity, specificity, a next step. Instead, most errors deliver the opposite: vague language written for the engineer who built the thing, not the person trying to use it.
A good error message names what happened, why it matters, and what to do next. Three sentences is enough. “Something went wrong” is not an error message. It’s an apology from a system that stopped paying attention.
The quality of error handling is a direct measure of how seriously a team took the user’s experience under pressure. Scott Andrew Alpaugh has seen integrations fall apart not because the logic was wrong, but because the failure states were invisible — silent drops, ambiguous rejections, retries that looked like success.
If you want to know whether a system was built with care, don’t watch it work. Watch it fail.
The error message is the product. It’s what ships when everything else goes sideways, which means it’s what gets used when it matters most. Design it like it counts.