Voice Uptime SLA: What It Covers and How to Read OneA voice uptime SLA is a contractual promise that your phone service will be available a stated percentage of the time, measured over a set window, with a defined remedy if the provider misses it. The percentage is the easy part. What the SLA counts as "voice," what it excludes and how a miss turns into a credit decide whether the number protects your business.

If you run phones across many sites, read the definition clauses before you compare nines. Two contracts that both say 99.99% can promise very different things.

TL;DR

• 99.9% allows roughly 43.8 minutes of voice downtime a month, 99.99% about 4.4 minutes and 99.999% about 26 seconds.
• The covered service definition matters more than the percentage: it states what must fail for the clock to start.
• Exclusions (maintenance, third-party circuits, your own equipment) can remove most real-world outages from the calculation.
• Credits are a partial refund on the bill, not compensation for lost calls. Redundancy and failover design are what keep calls flowing.
• Verify the SLA yourself with independent call testing and keep your own outage log, because most providers require you to claim the credit.

Table of Contents

• What a voice uptime SLA actually promises
• What the nines allow in a month
• What counts as downtime, and what does not
• Calling architecture changes what the SLA covers
• Service credits: what you get and what you do not
• When voice downtime becomes a regulatory matter
• Redundancy is what makes the number real
• How to verify your voice SLA yourself
• Questions to ask before you sign
• How California Telecom approaches voice reliability
• FAQ
• Sources

What a voice uptime SLA actually promises

An SLA has four moving parts: a service definition, a target, a measurement method and a remedy. Most disputes happen in the first and third.

The service definition. One provider may define availability as the ability to place and receive calls on the public phone network. Another may define it as the provider's own platform being reachable, which is a narrower thing. Microsoft's explanation of its Teams Phone SLA is a useful example of the first style: it describes the commitment as users being able to start a PSTN call or dial into conference audio, and it warns that equipment or services outside Microsoft's control fall outside the guarantee, including third-party PBX systems integrated into Teams. The lesson applies to any provider: the more of your call path that sits outside the provider's control, the less the headline number covers.

The target. Usually 99.9%, 99.99% or 99.999%, stated per month. Some contracts state it per year, which hides short, painful outages inside a generous annual budget.

The measurement method. Who measures, from where, and at what interval? A provider measuring from inside its own data center can report 100% while your callers hear busy signals because the carrier handoff failed.

The remedy. Usually a service credit, expressed as a percentage of the monthly fee for the affected service, with a cap.

What the nines allow in a month

The arithmetic is simple. Take an average month of about 30.4 days, or roughly 43,800 minutes, and multiply by the unavailable fraction.

Each extra nine cuts the allowance by a factor of ten. That is why the jump from three to four nines is a big engineering step: it moves you from tolerating a single long outage in a month to tolerating almost none.

Two cautions when you read these figures:

1. An allowance is not a ceiling. A 99.99% SLA does not stop a single outage from lasting two hours. It means the provider owes you a remedy when the month's total goes over the allowance.
2. Monthly and annual numbers are not interchangeable. A provider can post a strong annual average while a bad month still breaches the monthly target. Check which window the contract uses.

For a deeper look at what the top tiers demand in practice, see our write-ups on what 99.99% means and how to hit it and on whether five nines is worth the cost.

What counts as downtime, and what does not

The definitions section of a voice SLA usually answers three questions.

Which components are covered

Voice service is a chain: your handsets or softphones, your LAN and WiFi, your internet or private circuit, the provider's session border controllers and call-control platform, the carrier that terminates to the public network, and the numbering and emergency-call systems. An SLA typically covers the provider's own platform and sometimes the carrier handoff. It rarely covers your local network or a circuit bought from someone else.

When the same provider supplies the circuit and the voice platform, a single SLA can span the whole path. When you bring your own internet, you effectively own the weakest link. This is the main reason multi-site operators often prefer one accountable provider for transport and voice.

Voice quality versus availability

Availability asks whether calls can be made at all. Quality asks whether they sound acceptable. Some contracts treat them separately: a call can connect with jitter and packet loss so severe that nobody can understand the caller, and still count as "up." If call quality matters to you (a contact center, a clinic scheduling line), look for measurable quality commitments such as latency, jitter and packet loss thresholds, and see how SD-WAN tuning can fix voice quality.

Common exclusions

Read the exclusions line by line. The usual ones:

• Scheduled maintenance, sometimes with a monthly cap and a notice period, sometimes unlimited.
• Outages caused by your equipment, configuration changes or third-party software.
• Failures in circuits or carriers the provider does not control.
• Force majeure events.
• Emergency maintenance, which some contracts exclude and others count.
• Periods when you did not open a ticket or did not give access to a site.

Maintenance is the most valuable exclusion to negotiate. If a provider can take your voice platform down for hours with notice and none of that counts, the percentage is cosmetic.

Calling architecture changes what the SLA covers

The same brand name can sit on very different designs. Take Microsoft Teams Phone as an example, since many businesses run voice through Teams. Whether you buy a calling plan from Microsoft, connect your own carrier through direct routing, or use an operator-managed service changes who runs each component and therefore which part of the call path any SLA touches.

Microsoft states that its Teams Phone SLA of 99.99% covers Calling Plans, Phone System including Direct Routing and Voice Apps, and PSTN Audio Conferencing. The session border controller or carrier you attach to direct routing is a separate contract with its own terms. If that carrier is down, the platform SLA may be perfectly intact while your phones are dead.

So ask for the SLA of every layer, and ask who owns recovery when the layers disagree. If you are evaluating Teams-based calling for your sites, ask the provider to show you in writing which components its commitment includes. We cover our own hosted voice in UCaaS (Hosted PBX), and the same questions apply to any platform.

Service credits: what you get and what you do not

When a provider misses its target, the usual remedy is a credit against a future invoice. Typical features:

• Tiered percentages. The deeper the miss, the larger the credit, often capped at a share of the monthly fee for the affected service.
• You must claim it. Many contracts require you to submit a request within a short window after the month ends, with your evidence.
• Credits are the sole remedy. Some contracts say the credit is your only recourse, even for repeated failures. Look for a chronic-failure clause that lets you terminate without penalty after several misses.
• Credits are calculated on the affected service only, not on your whole account.

The core point: a credit never replaces a lost sales call, a missed patient call or a store that could not take phone orders for an hour. This is why what your contract really protects is a different question from what it refunds. The protection comes from the design: diverse circuits, carrier redundancy, automatic failover and tested recovery.

When voice downtime becomes a regulatory matter

For some outages, an SLA is not the only standard in play. In the United States, the FCC requires qualifying providers, including interconnected VoIP providers, to report network outages lasting at least 30 minutes that meet other thresholds through its Network Outage Reporting System. Covered 911 service providers face an even tighter notification duty to 911 call centers. Your own contract does not change those obligations, but it should say how the provider will tell you about an outage that affects emergency calling at your sites.

For multi-site businesses, emergency-call behavior during a failover deserves a specific line item: does the backup path still deliver the correct location information for each site? Our VoIP over LTE checklist walks through that for wireless backup.

Redundancy is what makes the number real

An SLA percentage describes a risk the provider accepts. Redundancy lowers the risk itself. The patterns that matter for voice:

1. Diverse transport. A primary fiber circuit plus a second path on a different carrier or technology, such as 4G/5G, so a single cut does not take down a site.
2. Multi-carrier trunking. SIP trunks with more than one upstream carrier and automatic rerouting when one fails. Our runbook for how to prove SIP trunk failover in seconds shows how to test it rather than trust it.
3. Platform failover. Geographically separate call-control instances, so one data center loss does not strand your phones.
4. Local survivability. A site keeps basic inbound and outbound calling when it loses its WAN link.
5. A disaster recovery plan. Governance, tests and a defined owner for recovery. See our guide to voice disaster recovery for enterprise IT.

The general engineering idea behind failover is to stop sending traffic to a failing dependency and resume once it recovers. Microsoft's architecture guidance on the circuit breaker pattern describes exactly that behavior for repeated calls to a failing remote service. The same logic, applied to carriers and circuits, is what separates a voice failure that lasts seconds from one that lasts hours.

If you are still weighing whether a second line or path is worth it, here is why businesses need redundant voice lines.

Healthcare and other always-on settings raise the stakes further. Guidance on medical phone service continuity shows how reliability expectations are framed when a missed call can mean a missed patient. Think the same way about any line where an unanswered call has a real cost.

How to verify your voice SLA yourself

Trusting the provider's dashboard alone leaves you unable to dispute anything. Build your own evidence.

Step 1: Define your own failure

Write down what "down" means for you: no inbound delivery, no outbound dialing, one-way audio, or call quality below a threshold. Compare that to the contract definition and note the gaps.

Step 2: Run independent test calls

Place automated test calls from outside your provider's network, at regular intervals, to your main numbers at each site, and record connect success, post-dial delay and audio quality. Outside-in testing sees what your customers see.

Step 3: Monitor the path, not just the platform

Track latency, jitter, packet loss and circuit status on each WAN link carrying voice. These metrics often warn you before calls fail. Our piece on SD-WAN SLA metrics to monitor lists the ones worth alerting on, and Netverge monitoring is the platform we use to watch them across sites.

Step 4: Keep an incident log

For every event, record start time, end time, affected sites, ticket number and how you learned of it. Providers dispute minutes. A timestamped log from two independent sources settles most disputes.

Step 5: File the claim on time

Calendar the claim window from your contract. If credits are not automatic, a missed deadline forfeits the remedy.

A clear process for matching contract language to measured data is the heart of SLA compliance work in any industry, and this SLA compliance playbook shows how operations teams turn commitments into measurable checkpoints.

Questions to ask before you sign

• What exactly must fail for the clock to start? Ask for the definition in plain language.
• Is the target monthly or annual? Who measures it, and from where?
• How many hours of scheduled maintenance are excluded, and with what notice?
• Does the SLA cover the circuit, the platform and the carrier handoff, or only one of them?
• Are voice quality metrics included, with numbers?
• What is the credit schedule and cap, and is the claim automatic?
• Is there a chronic-failure termination right?
• What is the escalation path at 2 a.m., and do I reach an engineer or a queue?

A provider that answers these directly and in writing is usually one that has thought about the failure modes. If you are choosing a partner for the engineering side, see how to evaluate outsourced network engineering providers.

Our perspective

We design voice SLAs backwards from the failure. Before we quote a percentage, we ask what a missed call costs you at each site, then build the transport, carrier diversity and failover to match. A number on paper is only as good as the path under it.

That is why we offer one provider, one bill and one engineer's number across voice and data. We source from 50+ carriers, our own engineers design and deploy each site, and a 24/7 U.S.-based NOC watches the service. Our SLAs are 99.99% on data and 99.999% on voice, and when voice and transport sit under one roof, there is no gap between two contracts for an outage to fall into.

Our advice: ask every provider to show you the failover working before you sign, and keep your own record of every minute. A good provider will welcome both.

Jim

How California Telecom approaches voice reliability

We run managed voice and the network beneath it for multi-location businesses. Our hosted voice sits on UCaaS (Hosted PBX), paired with dedicated fiber, wireless backup and managed SD-WAN so each site has more than one path. Our 24/7 U.S.-based NOC and Netverge observability give your team a single view of what is happening across your locations.If you want to compare your current voice contract with what a multi-carrier design would look like, talk to our engineers about your sites.

FAQ

Is 99.99% uptime good enough for business voice?

For most multi-site businesses, yes, as long as the SLA covers the whole call path and the design includes diverse circuits and failover. It allows about 4.4 minutes of downtime a month. Contact centers and emergency-critical lines may want the higher 99.999% tier.

Does an uptime SLA guarantee my phones will not go down?

No. It sets a measured target and a remedy, usually a bill credit. Outages still happen, so redundancy and tested failover are what protect your calls.

Do service credits cover lost business?

Usually not. Credits are typically a percentage of the monthly fee for the affected service, often capped, and many contracts name them as the only remedy. Look for a chronic-failure clause that lets you exit after repeated misses.

How do I check whether my provider met its SLA?

Run independent test calls from outside the provider's network, monitor latency, jitter and packet loss on each link, and log every incident with timestamps. Then compare your records with the provider's report before the claim window closes.

Sources

• Updated Microsoft Teams Phone SLAs reflect commitment to quality and continuity
• FCC Network Outage Reporting System (NORS)
• Circuit Breaker Pattern, Azure Architecture Center
• Medical Phone Service Continuity: A Practical Guide
• Installation SLA Compliance: A Practical Ops Playbook

Recommended

• SLA vs QoS: What Each Covers and When You Need Both
• SLA Credits vs Reliability: What Your Contract Really Protects
• Predict Outages: 4 SD‑WAN SLA Metrics Network Engineers Must Monitor
• 3 SD‑WAN Tuning Levers That Fix Voice Quality