Contact operations
The one application where demand is a plateau rather than a peak, which changes how the connectivity boundary has to be set.
- Plateau-aware sizing — Interval data rather than averages, because sustained load is not a peak.
- Partition isolation — Campaign traffic bounded in both directions.
- Investigable quality — Per-call metrics granular enough to act on.
- Same-day reporting — Latency low enough to inform tomorrow's roster.
Which boundary matters
The connectivity boundary, because sustained concurrency behaves differently from bursty concurrency. Pooling assumes peaks do not coincide; a campaign running all afternoon is not a peak that will pass.
The evidence boundary matters commercially. Degraded audio lengthens handling time, increases repeat contact and measurably reduces conversion, so a report of bad audio needs to be investigable rather than dismissible.
The control boundary matters for reporting latency, because staffing decisions are made on the same day the data describes.
How the layers are set
Campaign traffic is partitioned from corporate voice with its own floor and ceiling, so it can neither consume what the rest of the organisation needs nor be starved by it.
Capacity is modelled from historic interval data rather than daily averages, with headroom for recurring peaks, then reviewed against measured concurrency after go-live.
Per-call quality records are retained and reportable at a granularity that lets an agent-reported condition be traced to a specific segment.
The layers usually involved
In the order they normally get fitted here.
From historic interval data rather than averages or agent count. We identify recurring peaks, provision above the observed peak with headroom, then review against measured concurrency after go-live.
Yes, through partitioning at the connectivity boundary. Campaign traffic gets a guaranteed floor and an enforced ceiling, so neither side can starve the other.