Measure the routes before you commission another latency board
A polished p95 chart on untagged endpoints only accelerates the wrong debate.
Teams under pressure often skip route and version tags and jump to a new latency board. Weeks later, product and platform argue over different totals for the same API because every timing event still shares one opaque label.
Prefer route, caller, and dependency tags over widget names. Keep environment out of the route alias and in a property. Write allowed values the same day a new endpoint ships—or the path becomes a free-form dump.
In an API Latency Baseline Audit we spend early days on the inventory, not on chart cosmetics. Boards get sharper only after the latency map stops arguing with itself.