Noticeable delay on VoIP calls

Measure the experience and path

Record whether people talk over each other or hear delayed responses, and which site or route is affected. Round-trip ping latency is not identical to one-way mouth-to-ear delay. Codec processing, packetization and playout buffering also contribute.

Delay, jitter and loss are different

Look for queueing

Compare an idle call with a call during a controlled upload or other busy period. If delay rises sharply under load, inspect the actual bottleneck and queue management. A large nominal bandwidth figure does not prevent a saturated upload queue.

Check whether a VPN, distant PBX or unnecessary media relay changes the route. Do not remove a security boundary merely to shorten a path without assessing the full design.

Review endpoint processing

Inspect packet duration and jitter-buffer behavior where supported. Increasing a buffer can smooth uneven arrivals but adds delay. Transcoding and conference processing can add work on a PBX even when the network itself is stable.

Validate the trade-off

Compare sound quality and conversational delay after one justified adjustment. Test during peak load and with the actual remote locations. Keep both timing and packet-loss observations; reducing buffering too far can make audio worse while making the delay statistic look better.