Why Sub-50ms Edge Audio Latency Matters for Outbound Call Centers

The Latency Budget and Why the Edge Consumes It

Every outbound call has a fixed latency budget. ITU G.114 sets the conversational quality ceiling at 150 ms one-way. Exceed that and both parties begin to notice a delay — responses feel delayed, overlapping speech increases, and natural turn-taking breaks down. For a call center agent trying to engage a skeptical contact in a sales or collections conversation, any perceptible lag translates directly into lower conversion rates.

That 150 ms budget is consumed by four segments: local agent audio processing (~25 ms), your WAN path to the carrier edge (~15–40 ms depending on ISP and geography), the edge node itself (PoP to PSTN gateway), and the PSTN circuit to the contact's handset (~20–50 ms). The edge segment is the one your SIP carrier controls.

UnlimCall's edge nodes deliver under 50 ms PoP-to-PSTN across all 33 active markets. That leaves the WAN and PSTN segments as the variables. A carrier whose edge contributes 80–120 ms — common in legacy carrier architectures with centralized media processing — has already consumed most of the budget before the call reaches the public network. See network requirements for VoIP outbound calling for how to verify your own WAN contribution.

How Edge Architecture Determines Latency

Legacy VoIP carriers route all media through a centralized media proxy — often a single geographic cluster that handles traffic from all regions. A call from a US agent to a Canadian contact may traverse a media path that goes US → central proxy (somewhere in the Midwest) → PSTN gateway in Canada. If the proxy is 1,200 km from both endpoints, it adds 20–30 ms each way for propagation alone, before any processing overhead.

Distributed edge architecture places media processing at regional PoPs close to both the SIP trunk origination point and the PSTN gateway. The result is a shorter physical media path and a lower fixed latency contribution. For a 50-seat outbound floor running predictive dialing at a 3:1 dial ratio, every 10 ms reduction in edge latency translates to a materially more natural-sounding call — particularly on long calls (over 3 minutes) where conversation rhythm matters most.

UnlimCall's network edge is built with distributed PoPs across all 33 markets, keeping the media path local to the terminating PSTN infrastructure in each country.

Latency and Agent Cognitive Load

Agents do not think in milliseconds, but they feel the effects. On a call with 200 ms one-way delay (400 ms round-trip), an agent who asks a question waits 400 ms before the contact's response begins reaching them. That gap creates pressure to repeat the question, interrupting the contact. Contacts interpret interruptions as impatience or inattention.

Internal studies at outbound call centers have measured statistically significant drops in contact-to-appointment conversion rates when round-trip latency exceeds 350 ms versus 180 ms — roughly a 7–12% degradation in conversion on the same script, same agent, same list, different network path. The mechanism is the same one that makes satellite-phone calls feel awkward: the brain interprets the delay as social distance or disinterest.

Keeping total round-trip delay under 250 ms (125 ms one-way) is the engineering target for a call center that competes on conversion quality.

Measuring Your Current Edge Contribution

Before attributing poor call quality to your carrier's edge, isolate the measurement:

1. Use traceroute -n <carrier_sip_ip> to count hops and measure RTT to the SIP signaling address. SIP and media travel similar paths.
2. Send a SIP INVITE with an RTP loopback request (if your SBC supports it) and measure the RTCP Round-Trip Time field.
3. Pull RTCP-XR Delay Since Last SR (DLSR) values from your SBC logs on active calls to calculate the one-way media path delay.

If the RTT to your current carrier's edge exceeds 60 ms from your primary site, the edge is consuming more than 30 ms of your one-way budget before the call even reaches the PSTN. That is the threshold where a carrier change typically produces a measurable conversation quality improvement.

How Codec Choice Interacts with Edge Latency

Codec delay adds to edge delay. G.729 adds 5 ms of algorithmic processing versus G.711 on every encode/decode cycle. When your SBC and the carrier edge must transcode — because they negotiated different codec preferences — the transcoding adds another 15–30 ms. On a path already consuming 80 ms at the edge, a transcoding step can push total one-way delay above the 150 ms conversational threshold.

Configure your SBC to prefer G.711 in the SDP offer (as covered in the codec selection guide) to eliminate transcoding on the carrier-side. When codec and edge latency are both optimized, the only remaining variable is your WAN path.