Introduction
Most businesses pick an internet service provider based on the number advertised in bold on the plan page — 500 Mbps, 1 Gbps — and assume that number settles the question of whether the connection can handle a business phone system. It doesn't.
VoIP calls are far less demanding on raw bandwidth than the marketing numbers suggest, and far more sensitive to a handful of network characteristics that never appear on a consumer-facing plan comparison: symmetric upload speed, jitter, packet loss, and whether the connection supports basic traffic prioritization. A gigabit connection with poor jitter control can produce worse call quality than a modest business-grade line engineered for voice.
This guide covers what actually determines whether an internet service provider is VoIP-ready, the specific numbers worth asking for before signing, and how to test a connection's real behavior instead of trusting a sales sheet.
Why a "Fast" Plan Isn't the Same as a VoIP-Ready One
Bandwidth speed and voice call quality are only loosely related past a fairly low threshold. A single VoIP call uses a modest, fairly constant amount of bandwidth regardless of whether the underlying connection is rated at 100 Mbps or 1 Gbps — what actually determines call quality is how consistently that bandwidth is delivered moment to moment, not how large the theoretical ceiling is.
This is why a residential-grade gigabit connection sometimes produces worse call quality than a business-grade connection at a fraction of the advertised speed: the residential plan is optimized for burst downloads — large file transfers, streaming — while a business voice-ready connection is engineered for the steady, latency-sensitive, bidirectional traffic pattern a phone call actually generates.
Bandwidth Math: How Much You Actually Need Per Call
A single VoIP call typically consumes somewhere in the range of 80–100 Kbps for a high-quality codec, including overhead, though the exact figure depends on the codec in use. Sizing bandwidth for a business phone system means multiplying that per-call figure by your realistic peak concurrent call count — not your total headcount, and not your average call volume, but the number of calls actually happening simultaneously during your busiest hour.
- A 10-person office with 5 concurrent calls at peak needs roughly 500 Kbps dedicated to voice — a small fraction of what most modern connections offer
- A 50-agent contact center with 40 concurrent calls at peak needs roughly 3.5–4 Mbps dedicated to voice, still modest by modern connection standards
- The gap between these numbers and typical advertised plan speeds is large enough that bandwidth ceiling is rarely the actual constraint — jitter, packet loss, and prioritization almost always matter more once basic capacity is covered
This is the central point worth internalizing: for the overwhelming majority of businesses, raw bandwidth is not the bottleneck standing between a mediocre and an excellent calling experience. The remaining factors below are.
Further reading: sizing wireless failover capacity
Why Symmetric Bandwidth Matters More Than Download Speed
Many residential and some business internet plans are asymmetric — fast download, much slower upload — because that pattern suits typical consumer usage (streaming, browsing) where download dominates. A phone call is fundamentally bidirectional: your outbound audio consumes upload bandwidth at essentially the same rate your inbound audio consumes download bandwidth.
A connection with generous download speed but a thin upload path can handle receiving audio fine while degrading the audio you're sending, and callers on the other end experience that degradation without you necessarily noticing it locally. Confirming a plan's actual upload speed — not just its headline download number — is one of the simplest, most overlooked checks before committing a phone system to a connection.
Latency, Jitter, and Packet Loss: The Numbers That Actually Break Calls
These three metrics, not raw speed, are what separate a connection that handles voice cleanly from one that produces choppy, delayed, or garbled calls.
- Latency — the one-way delay data takes to travel; noticeable conversational lag typically starts becoming apparent above roughly 150ms one-way, and international or satellite-backed connections are more prone to crossing that threshold
- Jitter — variation in latency from packet to packet; even with acceptable average latency, high jitter causes audio packets to arrive out of order or unevenly spaced, which is what produces the choppy, robotic quality callers describe as "breaking up"
- Packet loss — data that never arrives at all; voice traffic tolerates a small amount gracefully through codec compensation, but loss climbing past roughly 1% starts producing audible artifacts, and anything higher makes calls difficult to sustain
An ISP that can't state typical jitter and packet loss figures for its network, or treats the question as unusual, is a signal worth taking seriously — these are standard network engineering metrics for any provider that takes business connectivity seriously.
Further reading: Cisco: Understanding Delay in Packet Voice Networks
QoS and Traffic Prioritization: Does the ISP Even Support It
Quality of Service lets a network prioritize latency-sensitive voice packets ahead of less time-sensitive traffic — a large file upload, a software update — sharing the same connection. Without QoS, a big upload from someone else on the network can starve an active call of the bandwidth and consistency it needs, even when the connection's overall capacity is more than sufficient on paper.
Not every ISP offers configurable QoS, and not every connection type supports it equally well — dedicated business fiber typically offers the most reliable QoS implementation, while some consumer and shared-infrastructure connections offer little to no meaningful prioritization control. Asking specifically whether and how an ISP supports QoS for voice traffic, rather than assuming any business-labeled plan includes it, avoids a common and expensive surprise after go-live.
Reading an ISP's SLA Like a Phone System Depends On It
Because it does. An SLA that only guarantees uptime percentage without addressing latency, jitter, or packet loss thresholds is protecting against the wrong failure mode for voice — a connection can technically stay "up" while degrading badly enough to make calls unusable, and a pure uptime SLA offers no recourse for that scenario.
A voice-aware SLA specifies acceptable ranges for latency and packet loss in addition to uptime, defines what happens (credits, escalation path) when those thresholds are breached, and commits to a real response time for service-affecting issues rather than a vague best-effort statement. Reviewing the SLA specifically for these voice-relevant terms — not just the headline uptime number — is worth the extra few minutes before signing a contract a phone system will depend on.
Further reading: top UCaaS features and uptime SLAs
Testing a Connection Before You Commit
Sales materials describe intended performance; testing reveals actual performance. Before committing a business phone system to a new ISP, running the connection through a real test period — routing live or simulated call traffic across it for at least a week, covering both quiet periods and peak usage — surfaces problems a spec sheet never will.
Free and low-cost network testing tools can measure latency, jitter, and packet loss directly against the connection under real conditions, and comparing those measured figures against the thresholds described above gives a concrete answer instead of a guess. A short test period before a multi-year contract commitment is a small cost against the alternative of discovering a connection's real jitter behavior only after staff start complaining about call quality.
Conclusion
The plan page number that most businesses use to choose an internet service provider — the advertised download speed — is close to irrelevant for voice call quality once a fairly modest bandwidth floor is met. What actually determines whether calls sound clean is symmetric upload capacity, low and consistent jitter, minimal packet loss, real QoS support for voice traffic, and an SLA that addresses those specifics rather than just uptime.
Asking for these figures directly, and testing a connection's real behavior before signing a long-term contract, is what separates a phone system built on a connection that happens to work from one built on a connection engineered to.
