Uptime Guarantee: What the Nines Allow and How Downtime Gets Counted

An uptime commitment says a service will be usable for at least a stated share of a measurement period. It is usually the most prominent number in a service level agreement and the least informative on its own, because three other things decide what it is worth: the period it is measured over, the definition of downtime, and the exclusions.

The percentage is nonetheless a reasonable starting point, because it tells you what the supplier is prepared to put in writing. A supplier declining to commit at all is telling you something. So is one committing to a level that would require substantial engineering investment to reach.

The arithmetic is straightforward: multiply the period length in minutes by one minus the target. A thirty-day month contains 43,200 minutes; a calendar year contains 525,600. Doing this conversion turns an abstraction into something you can plan around.

The pattern is worth internalising. Each additional nine cuts the allowance by roughly a factor of ten, and the engineering required to achieve it rises steeply. The gap between 99.9 and 99.99 per cent is not a marketing refinement; it usually means redundant infrastructure, automated failover and an operational discipline that costs real money.

The measurement period

A monthly measurement caps each individual month. An annual measurement caps the total but permits it to be concentrated. Two suppliers advertising the same percentage may therefore be offering quite different protection, and the shorter period is generally better for a customer because it limits the size of any single event.

The definition of unavailability

This clause decides how often the commitment engages at all. If unavailability means the entire service is unreachable for all users, then a service that is unusably slow, or broken for one geography, or missing its most important function, is fully available by definition. Broader definitions that include substantial degradation are considerably more useful and correspondingly harder to obtain.

Exclusions

Scheduled maintenance, force majeure, customer-side problems and third-party network failures are standard exclusions and are generally fair, since a supplier cannot commit to matters outside its control. The question is whether they are bounded. A maintenance exclusion with no cap on hours and no notice requirement removes an unspecified share of the period from the calculation before anything is measured.

Service credits refund a proportion of the fees for the affected period. For most businesses this is a small fraction of what an outage costs. A sales team unable to reach the system for a day loses conversations, follow-ups and momentum, and none of that is recovered by a credit against a subscription.

That is not an argument against credits. It is an argument for reading them correctly. Their function is to demonstrate that the supplier is willing to attach a consequence to its own commitment, which is a genuine signal. The operational response to a critical dependency is not a better clause but a plan: knowing what the team does during an outage, how they capture work that would otherwise be lost, and how it is entered afterwards.

A published status page with a full incident history tells you more about a service than a contractual number does. Look for incidents recorded that did not breach the commitment, since a page that only reports major failures is reporting selectively. Look for post-incident explanations that describe what happened rather than expressing regret. And look for consistency, because a status page that stayed green through an outage that users clearly experienced is worse than no status page at all.

A service level agreement is the wider document containing the availability commitment alongside support responsiveness and remedies. Service credits are the standard remedy. Recovery time and recovery point objectives describe how quickly service is restored after a serious failure and how much data might be lost, and they matter more than availability in a disaster scenario. An opt-out clause triggered by repeated failures is the escalation beyond credits.

Related pages on evaluating and relying on a CRM:features overview, what a CRM is, book a demo, compare alternatives, pricing, and CRM for small business.