Channel sizing based on measured concurrency instead of a headcount guess, in a report the support team runs themselves.
The problem
“How many channels do we need?” comes up at every carrier renewal, and the usual answer is a guess. Teams records every PSTN call, but the admin centre shows call lists, not how many calls were up at once, which is what a trunk is sized on.
What I did
- Wrote a script that reads the raw admin centre exports and rebuilds second-by-second concurrency for each trunk.
- Handled how channels are actually sold: one pool across a carrier’s SBCs, a limit per SBC, or a limit per operator.
- Reported peak and P95 use, an Erlang B recommendation, time over the purchased limit, and rejected calls with the load at the moment each happened. That last part separates a capacity problem from a network one.
- Produced one HTML file that can be emailed straight to a customer, and a guide so the team can run it without me.
Result
The team can now answer the sizing question from the customer’s own call data.