"99.99% guaranteed availability" reassures leadership. Nobody translates that into 52 minutes per year — or reads page 14 of the terms where planned maintenance, "exceptional" DDoS, and client DNS failures do not count. The day checkout is down two hours on a Friday, support points to an exclusion. The percentage was only a slogan.
An SLA is a limited contractual commitment, not a promise your business never stops. Understanding numbers and exclusions beats one more nine on the sales slide. Two hosts showing the same percentage may protect radically different things.
Translating nines into minutes
Base: 365 days × 24 h × 60 min = 525,600 minutes/year.
| Advertised SLA | Theoretical max downtime / year | Equivalent / month |
|---|---|---|
| 99% | ~87 h 36 min | ~7 h 18 min |
| 99.9% | ~8 h 45 min | ~43 min |
| 99.99% | ~52 min | ~4 min |
| 99.999% | ~5 min | ~25 s |
These figures assume every qualified minute enters the calculation — which contracts rarely cover fully.
What exclusions remove from the count
Announced maintenance is often excluded even when your site is offline. Client factor — disk quota, expired certificate — excludes the outage. Third parties, attacks beyond a threshold, and off-datacenter causes appear regularly. Two 99.9% contracts can have incomparable exclusion universes.
For real continuity, cross SLA with a disaster recovery plan and reliable status page. The reading an SLA guide details promise vs remedy.
SLA credits: real compensation
When breach is proven, the host often grants a credit on a future invoice: 5–10% of affected month, capped at 100% of the month — rarely more. Claim window 7–30 days. An €8 credit on an €80 VPS does not compensate €40,000 in abandoned carts.
Hero percentage sells seriousness; capped credits limit liability.
The climax: the percentage sells seriousness, not continuity
Compare on evidence via the directory and comparator — not the badge alone.
Decide and move forward without blind spots
Convert each advertised SLA into theoretical lost minutes, then list exclusions line by line in a document shared with the technical team. Verify measured scope: hypervisor, public HTTP, or entire application. Require a status page with incident history rather than a percentage with no method. Cross uptime SLA with support SLA — good network with bad human response leaves your team alone at night. Finally, calibrate business RTO and RPO independently of the displayed percentage.
Frequently asked questions
How much downtime for 99.9% per year?
About 8 h 45 min per year (525 minutes) if everything counts without exclusions. 99.99% ≈ 52 min; 99.999% ≈ 5 min. These are theoretical caps — real scope rarely covers your whole site.
Does the SLA cover my entire WordPress site?
Often not: scope limited to hypervisor, datacenter network, or control panel. Your app, DNS, or a plugin may be excluded.
Are SLA credits real compensation?
Generally 5–10% of monthly bill, capped, claimed within 30 days with proof. Not indemnity for lost revenue.
How do you compare two hosts on uptime?
Same scope, same exclusions, public status page, incident history — not just the homepage percentage.
Next time someone sells you 99.99%, ask: 52 minutes of what, excluded how, credited how much? A percentage without scope is not an SLA — it is decoration.