Contents
At a glance
- Category
- Business and buying
- Three required parts
- The metric, the target, and the remedy when it is missed
- 99.9% uptime
- Roughly 43 minutes of downtime per month
- 99.5% uptime
- Roughly 3.6 hours per month
- Read first
- The exclusions - maintenance, third parties and beta features are commonly carved out
- For a build project
- Agree fault response times, support hours, responsibility split and included maintenance
What an SLA actually commits to
A service level agreement is a written commitment about how a service will perform, with three parts that all have to be present to mean anything: the metric being promised, the target, and the remedy when it is missed. A document with the first two and not the third is a statement of intent.
The remedy is usually a service credit - a percentage of the monthly fee refunded - which is worth understanding clearly: it compensates you for the service you did not receive, not for the business you lost while it was down. If an outage costs you far more than the subscription, the SLA does not cover that gap and no small-business supplier will agree to a clause that does.
Reading uptime numbers
Uptime percentages are the headline and they compress a lot into a small number. 99.9% allows roughly 43 minutes of downtime per month; 99.5% allows around 3.6 hours; 99% allows about 7.3 hours. The step between them is larger than the decimal suggests.
Two questions matter more than the percentage. What counts as downtime - partial degradation, a slow API and scheduled maintenance are frequently excluded, which can make a headline figure considerably less impressive. And who measures it, since a supplier reporting on its own availability using its own definition is a weaker commitment than one with independent monitoring.
What to ask for in an automation project
For a build-and-hand-over project, uptime is largely the platform vendor's business rather than the builder's - your automation is only as available as n8n or Make is. What you should agree with a builder is different and more useful: response time to a reported fault, by severity; support hours and how to raise something outside them; who is responsible for what once it is handed over; and what maintenance is included versus billed.
That last one is where most disputes originate. An API change that breaks a workflow six months after handover is nobody's fault, and whether fixing it is included should be written down before it happens rather than negotiated afterwards - see handover.
SLAs you make to your own customers
Automation makes internal service commitments achievable, which is often the point. A promise to respond to every enquiry within fifteen minutes during business hours is difficult with people alone and straightforward with missed-call text-back and automated acknowledgement.
Before promising it, instrument it. Measure your current response time distribution at the 90th percentile, not the median, and promise something you hit consistently. A published commitment you miss regularly does more damage than no commitment at all.
What you should know
Read the exclusions before the percentage. Scheduled maintenance windows, third-party dependencies, force majeure and beta features are all commonly carved out, and the exclusions frequently matter more than the target. Also check how a claim is actually made - many require you to notice, report and claim within a short window, so a credit you never claim is a credit you do not get.
Note as well that an SLA is a commercial commitment rather than a technical control. It does not make anything more reliable; it establishes who bears the cost when it is not. Reliability itself comes from retries, queues, error handling and monitoring on your side.
SLA FAQ
What is an SLA?
A written commitment about service performance with three parts: the metric promised, the target, and the remedy when the target is missed. Without the remedy it is a statement of intent rather than an agreement.
How much downtime does 99.9% uptime allow?
Roughly 43 minutes a month. For comparison, 99.5% allows around 3.6 hours and 99% allows about 7.3 hours. The gap between those decimals is much larger in practice than it looks on a comparison table.
What does a service credit actually cover?
The service you did not receive - typically a percentage of the monthly fee - not the business you lost while it was down. If an outage costs you far more than the subscription, the SLA does not close that gap, and no small-business supplier will agree to a clause that does.
What SLA should I ask an automation builder for?
Not uptime, which mostly belongs to the platform vendor. Ask for fault response times by severity, support hours and out-of-hours escalation, a clear split of responsibility after handover, and exactly what maintenance is included versus billed.
Who fixes my automation when an API changes?
Whoever your agreement says, which is why it should be written down before it happens. An API change six months after handover is nobody's fault, and whether the fix is included, billed hourly or covered by a retainer is the most common source of disputes.
Should I offer my own customers a response-time SLA?
Only after instrumenting it. Measure your current response times at the 90th percentile rather than the median and promise something you hit consistently. A published commitment you miss regularly does more damage than making no commitment at all.
Primary source
NIST AI Risk Management Framework - governance and accountability
Checked August 2026. Anything priced or rate-limited moves, so verify against the source before you budget.
Want this built rather than explained?
Knowing what sla means is one thing; having it running in your business is another. We scope one workflow at a time, quote a fixed price in writing after a free consult, build it, and hand it over with documentation you own.
