Business Technology

Understanding Call Latency: What It Is and How to Prevent It on Business VoIP Calls

The one-way delay limits that the ITU standard and the major phone platforms publish, where the milliseconds come from, and the order in which to fix a lagging office phone system.

A small office network rack with tidy cabling

Call latency, the delay most people describe as call lag, is the time between a word being spoken and being heard. The major business VoIP platforms publish the same target as of October 2026: one-way delay of about 150 ms or less, jitter of a few tens of milliseconds and almost no lost packets. This article explains where delay comes from and how to remove it.

Call latency: the one-way, mouth-to-ear time it takes speech to travel from the person talking to the person listening on a phone call, counting the microphone, the codec, every network hop, the receiving buffer and the speaker. It is measured in milliseconds, and it is different from bandwidth, which measures how much data a connection can carry.

What does latency do to a phone call?

A small delay goes unnoticed. A larger one produces problems people recognize immediately.

  • Talk-over. Each side speaks into a pause that is really the other person's delayed reply, so both interrupt.
  • Echo. Your own voice can return late enough to be heard as an echo. Cisco's reference on packet voice delay notes that echo cancellers become necessary once one-way delay passes about 25 ms (Cisco, Understanding Delay in Packet Voice Networks).
  • Robotic or clipped audio. This is usually jitter, not steady latency: the receiving phone's buffer runs dry or overflows and the audio breaks up.
  • Dropped calls. When signaling packets arrive too late, the call itself fails.

How much latency is acceptable?

The reference is ITU-T Recommendation G.114, the international standard on one-way transmission time. It recommends that one-way delay not exceed 400 ms for general network planning, and states that if mouth-to-ear delay can be kept below 150 ms, most applications, speech and non-speech, experience "essentially transparent interactivity" (ITU-T G.114, 2003 edition, still listed as in force by the ITU).

The business platforms turn that into targets you can test against. The table lists what each vendor publishes for its own service.

Source One-way latency Jitter Packet loss Audio bandwidth per call
ITU-T G.114 (planning standard) Under 150 ms preferred, 400 ms ceiling Not specified Not specified Not specified
Nextiva Under 150 ms ping to its servers Under 15 to 20 ms No figure given 100 kbps
RingCentral Under 150 ms (under 300 ms two-way) Not over 40 ms Not over 1% No figure given
Microsoft Teams (to the Microsoft edge) Under 60 ms round trip for "optimal" Under 3 ms optimal, over 30 ms poor Under 0.5% average optimal 58 kbps recommended, 76 kbps best
Zoom Phone Not published on its requirements page Not published Not published 60 to 100 kbps

Sources: Nextiva networking guidelines, RingCentral quality of service guidelines, Microsoft media quality and network connectivity performance, Microsoft Teams bandwidth requirements and Zoom system requirements. Microsoft's targets are stricter because they cover only the segment from your office to the nearest Microsoft edge, not the whole call.

Where do the milliseconds come from?

Latency is a budget spent in pieces, and most of the pieces are inside your own building or your provider's design, not on the public internet. Cisco's delay reference breaks a packet voice call down this way (Cisco):

  • Codec and packetization. The phone compresses speech and waits to fill a packet before sending it. Cisco's rule of thumb is a packetization delay of no more than 30 ms, with 20 ms the common setting for both the compressed G.729 and the uncompressed G.711 codecs (Cisco).
  • Queuing. The time a voice packet waits behind other traffic in a router or switch. This is the part quality of service controls, and the part that grows during a large upload or a video meeting next door.
  • Propagation. Distance at the speed of light in fiber, about 6 microseconds per kilometer in Cisco's estimate.
  • Jitter buffer. The receiving phone deliberately holds packets so that variable arrival becomes steady playback. The bigger the jitter, the bigger the buffer, and every millisecond of buffer is added latency.

Why does latency happen in a small office?

  • Congestion on the uplink. Upload capacity is usually smaller than download, and voice leaves the building on the upload side. A cloud backup, a large file sync or a few video meetings can fill it.
  • No prioritization. Without quality of service, voice packets wait in the same queue as everything else. RingCentral marks voice media at DSCP 46 (EF) and SIP signaling at DSCP 26 (AF31) (RingCentral), and Microsoft says QoS should be implemented on every segment of a managed network (Microsoft Learn).
  • Wi-Fi. A wireless phone or softphone contends for airtime with every other device on the access point, and retransmissions show up as jitter. Microsoft's guidance is to use Wi-Fi Multimedia (WMM) prioritization, prefer the 5 GHz band, steer dual-band clients to it and keep neighboring access points on non-overlapping channels (Microsoft Learn).
  • Consumer routers and helper features. Nextiva asks customers to disable SIP ALG (some routers use a different label for it) because it rewrites signaling and causes one-way audio, deregistrations and dropped calls, and to avoid double NAT, meaning two routers in series (Nextiva). RingCentral gives the same SIP ALG advice.
  • Wireless internet service. Nextiva lists 5G and satellite internet providers as unsupported because of packet loss, and RingCentral notes that geosynchronous satellite links add delay by their nature.
  • VPNs. Routing voice through a VPN adds encryption and often a detour. Microsoft recommends a split tunnel so that real-time media bypasses the VPN (Microsoft Learn).

How do you prevent call latency?

Measure first. Every vendor above publishes a test tool or a dashboard, and Microsoft's advice on method is the useful one: sample for at least a week, every ten minutes, and compare the 90th percentile, not the average, against the target (Microsoft). Then work through the list in this order, cheapest first.

  1. Size the circuit for the busy hour. Count simultaneous calls, multiply by the per-call figure your platform publishes (about 100 kbps for Nextiva, 60 to 100 kbps for Zoom Phone), then add the video meetings, backups and everything else on the uplink. Our VoIP bandwidth calculator does the arithmetic. If the busy-hour total is near the upload capacity, upgrade the circuit or add a second one.
  2. Turn on quality of service end to end. Mark voice as expedited forwarding and give it a priority queue on the firewall, the switches and the Wi-Fi. On a Cisco Meraki MX, a traffic shaping rule for the VoIP and video conferencing category can be set to high priority and tagged with a DSCP value, and the rule only works as intended if the uplink throughput setting matches the real circuit (Cisco Meraki traffic shaping). Fortinet and Ubiquiti equipment has equivalent controls; our networking service configures them.
  3. Wire the desk phones. Nextiva's guideline is that all phones should be hardwired on Cat5e or Cat6 cabling for best quality (Nextiva). Put the phones on their own voice VLAN so that a misbehaving PC never reaches them; structured cabling that reaches every desk makes this routine.
  4. Fix the router configuration. Disable SIP ALG, remove any second router in the path, open the provider's port ranges and keep firmware current. Replace a consumer router with a business firewall that can enforce the priority queue.
  5. Make Wi-Fi fit for voice where it must carry it. Enable WMM, prefer 5 GHz, plan access point placement for coverage without overlap, and keep softphones on a wired dock when the person is at a desk all day.
  6. Add a failover path for offices that cannot lose the phones. A second circuit with automatic failover, or SD-WAN that steers voice to whichever link is performing, removes the worst case; when SD-WAN is worth it covers the decision.
  7. Monitor continuously. Call quality dashboards on the phone platform and uplink utilization graphs on the firewall show congestion building before users complain.

When is it the provider, and when is it the internet?

If the measurements at your edge are clean but delay appears beyond it, there are three remaining suspects. The provider's point of presence may be far away, so ask where your calls are terminated and whether a closer region is available. The internet provider's routing may be inefficient; a traceroute shows that. Or the circuit may be oversubscribed in your neighborhood at busy hours, which shows up as loss and jitter that follow the clock; business internet in Irvine: fiber vs cable vs fixed wireless compares the circuit types.

Key takeaways

  • Keep one-way delay under about 150 ms and jitter to a few tens of milliseconds; the ITU standard and the major platforms agree on the order of magnitude.
  • Speed and latency are different measurements. Fast fiber with no prioritization can still produce laggy calls.
  • Most of the delay budget is spent in your own building: queuing, Wi-Fi contention, consumer routers and VPN detours.
  • Measure during the busy hour, compare the 90th percentile with the vendor's target, then fix in order: bandwidth, QoS, wiring, router settings, Wi-Fi, VPN, failover.

How Telxpress helps

Telxpress designs the networking and connectivity under a business phone system so that voice has its own priority queue, its own VLAN and a wired path to every desk, and supports the phone systems and VoIP that run on it. If calls already lag, we measure latency, jitter and loss at your edge before we recommend anything, and we say plainly when the fix is the provider's. Book a consultation to get the measurement started.

  • legacy
  • voip
  • call-quality
  • latency

Questions

Questions this article answers

How much call latency is too much?

The international planning standard says most applications feel transparent when one-way mouth-to-ear delay stays under 150 milliseconds, and it sets 400 milliseconds as the ceiling for network planning. Business VoIP providers publish similar targets, usually 150 milliseconds or less measured to their servers. Above that, people start talking over each other.

How is latency different from jitter and packet loss?

Latency is a constant delay: everything arrives late by the same amount, so people interrupt each other. Jitter is variation in that delay, which the phone smooths with a buffer that adds more delay, and which sounds robotic when it overflows. Packet loss is audio that never arrives, heard as gaps. Each has different causes and fixes.

Why does the office have fast internet but laggy calls?

Speed and delay are different measurements. A fast connection can still queue voice packets behind a large upload, run voice through a consumer router with no prioritization, or carry calls over a congested Wi-Fi band. Measure latency, jitter and loss to the provider during a busy hour before assuming the circuit is the problem.

Can a cheap router cause latency?

Yes. Consumer routers often lack real quality of service, struggle under many simultaneous connections and run old firmware. Some also enable a SIP helper function by default that the major providers ask you to switch off. A business firewall that prioritizes voice, paired with managed switches, removes a common source of delay and one-way audio.

Is latency the provider's fault or ours?

It can be either, which is why diagnosis starts with measurement at your edge. If delay or jitter is high before traffic leaves your network, the fix is local: prioritization, cabling, Wi-Fi, equipment or bandwidth. If your network is clean and the delay appears between your edge and the provider, the circuit or the provider's routing is the issue.

What causes call lag on business phone calls?

Call lag is latency the caller can hear. The usual causes sit inside the office: voice packets queuing behind uploads and video on a small uplink, no quality of service on the firewall and switches, softphones on crowded Wi-Fi, a consumer router rewriting call signaling, and voice routed through a VPN. Distance to the provider adds the rest.

Next step

Want This Applied to Your Business?

Book a free 30-minute AI readiness consultation. You'll leave with three AI opportunities specific to your business, whether you work with us or not.

Or call 24/7: (949) 861-4500

Last reviewed: . Published by Telxpress, Irvine, CA.