Measure relief, not only saturation
Capacity metrics that ignore time-to-relieve leave planners flying blind after the first congestion alert.
Saturation counts tell you something filled. Relief metrics tell you whether traffic — and the team — can return to a known-good headroom. Application analytics should emit relieve markers: when utilisation dropped below threshold, when backlog drained, when the alternate path took load.
Without those markers, incident reviews invent timelines from chat. With them, you can compare congestion events on mean time to relieve and spot processes that routinely stall.
Instrument relieve events with the same care you give peak events. Pair them in reviews. Capacity planning improves when relief is a first-class signal, not an afterthought in a spreadsheet.