What this page is
A number we commit to, and the way to check it. We measure our services continuously, we publish the target, and when we miss it you get a credit without having to argue.
It is not a promise of zero downtime. No honest provider can promise that. It is a promise of measurement: a target, probes, written exclusions, and a credit if the target is missed.
What is covered
- Web hosting (sites, applications, databases)
- Email and productivity
- Online backups
- File synchronisation
- Dedicated servers we manage
Dedicated and custom servers: two options are required for the SLA to apply — monitoring and alerting, and local VM backups. Without them, we can neither detect the failure nor restore the server.
What is not covered
- Domain names — a registrar service
- 1Password — a third-party service, resold by us
- eM Client — software installed on your machine, which does not depend on our servers
- Security certificates — once installed, there is no service left to maintain
The target: 99.9% per month
99.9% over a 30-day month is 43 minutes of downtime. Not 99.99%: a target that cannot be measured honestly protects nobody.
Measurement is continuous, 24 hours a day, 7 days a week. Maintenance windows do not remove hours from the calculation: only the actual interruption counts.
How we measure
Probes query each service once a minute: the site home page, an email connection test, the required ports. Response time is compared with the service's historical trend, because a busy server is not a dead server.
Unavailability starts when the probe sees it, not when you call. That is what makes the calculation verifiable: nobody has to take anyone's word for it.
Maintenance
Security updates come before everything else. When a vendor fixes a flaw, the fix is applied quickly — that is what protects your data.
| Window | When (Eastern) | For what |
|---|---|---|
| Primary | Sunday 5 p.m. → Monday 5 a.m. | Updates and replacements that do not affect availability |
| Security | Every evening 5 p.m. → 5 a.m. next morning | Fixes that would affect availability if postponed; daily backups run in this window |
| Critical security | Any time, without notice | A flaw already exploited or a door left open; we inform you afterwards |
Inside these windows, routine fixes are not announced, and the interruption usually lasts 0 to 30 minutes. Maintenance expected to take longer is planned and announced in advance.
Those interruptions do not count as unavailability, up to 1% of the month's time — about 7 hours. Beyond that threshold, the excess counts and may earn a credit. They protect the service and keep it current, and they are kept as short as possible. A service that stays up during the window counts, of course, as available.
Return to service
After a hardware or virtualization failure, we target a return to service within 24 hours of the unavailability detected by our probes. In most scenarios it is faster — recovery scenarios are tested regularly.
What is excluded
- Any interruption caused by maintenance — fixes, updates, planned work — up to 1% of the month's time (about 7 hours); beyond that, the excess counts
- Force majeure: power or telecommunications failure, disaster, labour dispute, government decision
- Failure at your end, at your access provider, or on a network we do not own
- Third-party services: external DNS resolution, registrar, certificate authority
- Client acts or omissions, including a configuration error
- Suspension for non-payment or prohibited use
- A major security incident, including ransomware, despite the measures taken
- Hardware or software we do not manage
Credits
| Measured availability (month) | Credit |
|---|---|
| 99.9% and above | no credit |
| 98.0% to 99.89% | 25% of the monthly fee for the service |
| 95.0% to 97.99% | 50% of the monthly fee |
| below 95.0% | 100% of the monthly fee |
Cap: 100% of the monthly fee for the affected service. The credit is applied to the next invoice. It cannot be converted to cash or transferred to another account, and it is calculated on the amount actually paid for the period.
How to claim
- Open a ticket from your client area, within 30 days of the end of the month concerned.
- Name the affected service and, if you have them, the times observed.
- We answer within 30 days and apply the credit to the next invoice.
You do not have to convince us: our probes hold the same data you do. If the target was missed, the credit is owed.
The credit is the remedy provided for a failure to meet the availability target. It does not remove any right the law gives you.