I ping the login, a cheap query, and the last successful job time. If any of those fail during operating hours, Slack gets a quiet message, not a marketing-status incident.
Uptime for a lookup app is as real as uptime for a storefront. The revenue is just one step downstream.
Synthetic checks from inside the building
A monitor on the public internet can miss the warehouse Wi-Fi dying. A phone on the floor that hits the health endpoint once an hour is a better canary.
Dependencies on a sticky note
Supabase, a WMS, an OCR key. When the app is "down," I want to know which one. A status field per dependency beats a generic spinner.
