People are stopped by errors that explain nothing at all

• Give each failure path its own message rather than one shared string
• Say whether retrying will help, and after how long
• Start with the sign-in failure that survives a reinstall

Why it might not workSpecific errors leak how a system works, which is why teams default to one vague string.

• Audits the failure paths a team never sees itself
• Delivered as a rewritten set rather than a report

Why it might not workIt is the kind of work teams believe they can do themselves, and usually postpone.

• Compare completion after a specific message versus a shared one
• Measure retries before abandonment

Why it might not workThis needs production traffic, so it can only run inside a company that already has the problem.

• Collect the same vague string across several products
• Use the reviews where reinstalling changed nothing

Why it might not workError writing has a small literature already; the angle is what the vagueness protects.