99.9% vs 99.95% Uptime: What an SLA Covers
What 99.9%, 99.95% and 99.99% uptime mean in minutes of downtime, what a call center SLA can and cannot cover, and the questions to ask a vendor first.
Kurze Antwort
A 99.95% uptime SLA allows about 21.6 minutes of downtime in a 30-day month and about 4 hours 23 minutes in a year. 99.9% allows 43.2 minutes a month, and 99.99% about 4.3 minutes. A call center SLA normally covers the vendor's own platform only. Your carrier, internet connection and CRM sit outside it, so ask what is measured, what is excluded and how credits work.
Uptime percentages look close together and behave very differently. The gap between 99.9 and 99.95 is a factor of two in downtime. The more useful question for a contact center is not the number but the sentence behind it: up for whom, measured where, and covering which part of a call?
This guide turns the percentages into minutes, explains what "up" should mean when you run agents and live calls, lists the parts of a call a platform SLA cannot cover, and gives you questions to put to any vendor. The arithmetic needs no external statistics.
What does 99.95% uptime mean in minutes?
Downtime allowance equals the total minutes in the period multiplied by the unavailable share. A 30-day month has 43,200 minutes and a 365-day year has 525,600.
| Betriebszeit | Allowed downtime per 30-day month | Allowed downtime per 365-day year |
|---|---|---|
| 99.9% | 43.2 minutes | 525.6 minutes (8 hours 45.6 minutes) |
| 99.95% | 21.6 minutes | 262.8 minutes (4 hours 22.8 minutes) |
| 99.99% | 4.32 minutes | 52.56 minutes |
Two readings of these numbers matter. First, a month of 21.6 minutes can be one outage or ten, and the SLA may not distinguish. A single 20-minute outage during your busiest campaign hour costs far more than 20 quiet minutes at 3 a.m. Second, the percentage resets each period. A bad week early in a month uses the whole allowance for that month.
Worked example: why the chain is weaker than any one link (illustrative)
When a call needs every part of the path to work, the availabilities multiply. Take three example figures: a platform at 99.95%, a carrier at 99.9% and an internet line at 99.9%. These are example inputs, not anyone's published commitments.
- 0.9995 x 0.999 x 0.999 = 0.99750, or about 99.75% for a call that needs all three.
- 0.25% of a 30-day month (43,200 minutes) is roughly 108 minutes, against 21.6 minutes for the platform alone.
- Most of that gap comes from the parts the platform SLA does not cover, which is why the next two sections matter more than the headline number.
The multiplication assumes the failures do not overlap. Real outages sometimes overlap and sometimes do not, so treat it as a rough ceiling on what your end-to-end experience can be, not a forecast.
Worked example: when the 21.6 minutes lands (illustrative)
Suppose a team's busiest campaign runs two hours a day. A single 21.6-minute outage inside that window removes about 18 percent of that window's calling time (21.6 divided by 120), while the same outage overnight removes none of it. The SLA treats both identically. If your volume is concentrated, ask whether the vendor schedules maintenance outside your peak and how much notice you get.
What does "up" mean for a contact center?
A server answering a health check is not the same as a working contact center. Define "up" by what your people and customers experience. A practical definition has four tests:
- Agents can log in and reach their desktop.
- Calls connect, inbound and outbound, with audio in both directions.
- Calls are recorded and saved, where recording is required.
- Reports and records are readable, so supervisors and auditors can see what happened.
Many SLAs define availability around a narrower technical test. Neither approach is wrong, but you should know which one you are buying. If the definition is narrower than your four tests, the percentage tells you less about your agents' day.
Which parts of a call can a platform SLA not cover?
A call crosses several services, and a platform SLA covers only the one the vendor operates. If you Bringen Sie Ihren eigenen Anbieter mit., the voice path is shared between systems with different owners:
| Part of the path | Who runs it | Where its commitment comes from |
|---|---|---|
| Contact center platform | The platform vendor | The platform SLA |
| SIP trunk and carrier network | Your carrier (under BYOC) | Your carrier agreement |
| Office or home internet | Your internet provider | Your internet contract |
| CRM and other integrations | Their own vendors | Their own terms |
When a call fails, work out which row failed before blaming the platform. A call trace helps: on the Enterprise plan, the Betriebszentrale shows what the platform did with a single call, so you can tell a routing problem from a carrier problem. For trunk-side problems, see SIP trunk failover design.
Questions to ask about measurement, exclusions and service credits
Put these to any vendor, ours included, and ask for the answers in the agreement rather than on a web page:
- What exactly counts as downtime? Is a degraded service, such as calls connecting with no audio, counted?
- Who measures availability, from where, and how often?
- Over what period is the percentage calculated: a month, a quarter, a year?
- Which planned maintenance is excluded, and how much notice is given?
- Which components are outside the SLA?
- What is the remedy when the commitment is missed, and how do I claim it?
- Does the SLA apply to my plan, or only to some plans?
- What happens to recordings and data if there is an outage?
The answers are contract terms, so this guide does not say what is normal in the market. Use the list when you compare vendors, together with the Contact-Center-RFP-Checkliste.
How DialerBee handles it
The facts below are from our platform reliability page:
- Uptime SLA on Enterprise. The Enterprise plan carries a 99.95% uptime SLA, written into the agreement. The SLA is a feature of the Enterprise plan, so check the Preisseite und die enterprise solution page for what each plan includes.
- Two FreeSWITCH media nodes. Media runs on two nodes rather than one.
- Agent WebSocket failover. The agent desktop holds a WebSocket to a media node, and if that node goes away the client fails over to the other node at reconnect.
- Nightly off-site backups. The database is backed up nightly and the backups are held off-site.
Failover at reconnect means an agent may see a brief reconnect, not nothing at all. Ask any vendor how its failover behaves for a live call in progress, not only for the next login. Our cloud versus on-premise comparison sets this against running the stack yourself.
Designing around the gaps
The pieces an SLA cannot cover are the ones you can design for:
- Trunk failover. Keep a second route or carrier for outbound and a plan for inbound DIDs. See SIP trunk failover design.
- A second internet line for the office, or home agents on separate connections, so one provider outage does not stop all agents.
- A written outage plan that says who decides, who tells customers and what happens to callbacks. Use the RFP-Checkliste to ask the vendor, and keep your own copy of the answers.
- Your own monitoring. A test call every few minutes tells you about an outage before agents do.
When something fails, who do you call first?
An SLA only helps if you can tell which provider to contact. Keep this table next to your supervisors and fill in the names and numbers for your own setup.
| What agents report | First thing to check | Likely owner |
|---|---|---|
| Nobody can log in or the desktop will not load | Is it every agent or one site? Does another website load on the same connection? | Platform, or your internet provider |
| Logged in, but outbound calls fail to connect | Trunk registration and carrier status | Your carrier |
| Calls connect but audio is choppy for one agent | That agent's connection and headset | Your internet provider or the agent's setup |
| Calls work, but customer records do not open | The CRM's own status | CRM vendor |
If the trunk side is the problem, SIP trunk failover design explains how to give calls a second route.
How to keep your own availability record
Do not rely on the vendor's number alone. A short log lets you check an SLA claim against what your team lived through:
- Record the start and end time of every outage or serious degradation, in one time zone.
- Write what agents could not do: log in, place calls, receive calls, open records.
- Note which provider you believe was responsible and what evidence supports it, such as a call trace or a carrier notice.
- Add up the minutes per month and compare with the allowance in the table above.
- Keep the log for as long as your agreement says claims may be made.
The habit also tells you whether your real problem is the platform or one of the links around it. For many teams it turns out to be the link nobody was monitoring.
Häufig gestellte Fragen
What does 99.95% uptime mean in minutes?
In a 30-day month there are 43,200 minutes, and 0.05 percent of that is 21.6 minutes of allowed downtime. Over a 365-day year (525,600 minutes) it is 262.8 minutes, or about 4 hours 23 minutes. The figure only means something once the agreement defines what counts as down and how it is measured.
How much downtime is 99.9% uptime per month?
99.9 percent allows 0.1 percent downtime. In a 30-day month of 43,200 minutes that is 43.2 minutes. Over a 365-day year it is 525.6 minutes, which is 8 hours 45.6 minutes. That is exactly twice the allowance of 99.95 percent.
Does a call center SLA cover my SIP trunk or carrier?
Not automatically. A platform SLA covers the platform the vendor runs. If you bring your own carrier, the carrier's line, your office internet and your CRM are separate services with separate commitments, or none. Read the definition of the covered service and ask in writing which parts of the call path it includes.
What are service credits in an SLA?
Service credits are the remedy an agreement may offer when availability falls below the commitment, usually a reduction on a future invoice. Whether the agreement has them, how large they are and how you claim them are contract terms, so ask for the wording before signing rather than assuming.
What should I ask a vendor about its uptime SLA?
Ask what counts as downtime, how and where availability is measured, over what period, which planned maintenance is excluded, what the remedy is and how you claim it, and which components the SLA does not cover. Also ask how the platform behaves when one node fails, because that is what your agents will notice.
Which DialerBee plan has an uptime SLA?
The Enterprise plan carries a 99.95 percent uptime SLA, written into the agreement. Media runs on two FreeSWITCH media nodes, the agent desktop fails over between them at reconnect, and the database is backed up nightly with backups held off-site.
Quellen
- DialerBee platform reliability page, /platform-reliability/ (geprüft am 19.11.2026)
- Downtime figures are arithmetic on a 30-day month (43,200 minutes) and a 365-day year (525,600 minutes); no external statistics are used.
Weiterführende Lektüre
- Plattformzuverlässigkeit
- Vertrauenszentrum
- SIP Trunk Failover Design
- Operationszentrum
- Cloud-Dialer vs. On-Premise
- Contact-Center-RFP-Checkliste
Haftungsausschluss: This article is general information, not legal advice. Service terms depend on the agreement you sign, and the Enterprise SLA wording in your contract governs. Facts about DialerBee reflect the platform reliability page checked on 2026-11-19.
Weiterführende Lektüre: Call Center Business Continuity Plan Template
Verwandte Artikel
AI Voice Agent Latency: What Callers Notice
9 Minuten Lesezeit
TechnologieArabic-English Code-Switching in Voice AI and Transcription
9 Minuten Lesezeit
IndustrieRetention Calls: A Save Desk Playbook
11 Minuten Lesezeit
IndustrieDealer Recall and Service Reminder Calls
11 Minuten Lesezeit
Bereit, DialerBee in Aktion zu sehen?
Fordern Sie eine 15-minütige Live-Demo an oder starten Sie eine 14-tägige kostenlose Testphase – ohne Folien, ohne Verpflichtung. Das Telefonieren beginnt, sobald Ihr eigener SIP-Trunk oder Ihre TK-Anlage verbunden und freigegeben ist.
14-tägige kostenlose Testphase · kostenloses KI-Guthaben · keine Kreditkarte · 11 Sprachen · BYOC · Kontrollen zur Unterstützung der Compliance