Almost nobody actually needs custom software. What they need is for a specific, recurring problem to stop costing them time or money, and sometimes the honest answer is a better spreadsheet, not a six month build.
The real signal is repetition, not frustration
A process that is annoying once is not worth building software for. A process that is annoying every single day, done by more than one person, and prone to the same mistake each time, usually is. If nobody can describe the exact moment things go wrong, it is too early to scope anything.
Off the shelf tools fail quietly, not loudly
The usual failure mode is not that a generic tool cannot do the job, it is that it does eighty percent of the job and the missing twenty percent gets patched with a spreadsheet, a WhatsApp group, and someones memory. That patchwork is invisible until the one week the person holding it together is unavailable.
What we actually check before recommending a build
Before we write a scope, we ask what happens today, by hand, and how often it breaks. If the honest answer is "rarely, and it is fine," we say so, even though a rebuild would be the easier sale. If the pattern repeats daily across a growing team, that is usually the point custom software earns its cost back.