Average utilisation hides the paths that actually fill
A calm fabric average can mask a single saturated uplink if your application analytics never see path-level capacity.
Fabric averages feel reassuring. They are also easy to misread. If application analytics for network infrastructure capacity only roll up site-wide means, a single congested path can starve a critical sync while the board stays green.
Start by asking which application flows must complete under peak load. Then check whether those flows emit utilisation events your capacity metrics actually consume. Gaps between planning sheets and operational pain usually show up as latency tickets without matching capacity alerts.
A practical fix is to pair fabric averages with path-peak indicators — for example, 95th percentile utilisation on named uplinks — and review both in the same weekly capacity pack.