SaaS · 9 min read ·

SaaS Uptime SLAs: What "99.9%" Really Means (and What to Negotiate)

A 99.9% uptime SLA allows about 43 minutes of downtime a month, or roughly 8.7 hours a year, and the usual remedy is a small service credit. Check exclusions, how uptime is measured, how to claim credits and whether chronic failure lets you terminate.

Short answer: a "99.9% uptime SLA" sounds airtight, but in practice it permits the vendor about 43 minutes of downtime per month, or roughly 8.7 hours per year, and the remedy when they miss it is usually a small service credit on your next invoice, not real compensation. SLAs are one of the most-marketed and least-scrutinized parts of a SaaS contract. The number on the brochure feels reassuring; the fine print in the actual SLA is where the protections live or die. Here is what uptime percentages actually mean, how SLAs typically pay out, and what to negotiate so the SLA does something other than make you feel good.

Key takeaways

  • 99.9% uptime still allows roughly 8–9 hours of downtime a year; 99% allows over three days.
  • Exclusions for maintenance and third-party failures quietly enlarge the allowed downtime.
  • Service credits rarely match your real losses and usually must be claimed by you.
  • For critical systems, negotiate a higher SLA and a termination right for chronic failure.

Translating uptime percentages to real downtime

Uptime percentages sound similar but represent very different real-world tolerances. Here is what each common SLA actually permits the vendor:

  • 99% uptime = up to 7.2 hours of downtime per month, about 87.6 hours (3.65 days) per year.
  • 99.5% uptime = up to 3.6 hours per month, about 43.8 hours per year.
  • 99.9% uptime ("three nines") = up to ~43 minutes per month, about 8.76 hours per year.
  • 99.95% uptime = up to ~22 minutes per month, about 4.38 hours per year.
  • 99.99% uptime ("four nines") = up to ~4.3 minutes per month, about 52.6 minutes per year.
  • 99.999% uptime ("five nines") = up to ~26 seconds per month, about 5.26 minutes per year, rare outside top-tier infrastructure providers.
Downtime each common SaaS uptime percentage allows
Uptime commitmentAllowed downtime per monthAllowed downtime per year
99%Up to 7.2 hoursAbout 87.6 hours (3.65 days)
99.5%Up to 3.6 hoursAbout 43.8 hours
99.9% (three nines)About 43 minutesAbout 8.76 hours
99.95%About 22 minutesAbout 4.38 hours
99.99% (four nines)About 4.3 minutesAbout 52.6 minutes
99.999% (five nines)About 26 secondsAbout 5.26 minutes

The exclusions that quietly enlarge the allowance

The uptime percentage is the headline, but most SLAs exclude entire categories of downtime from the calculation. Common exclusions include scheduled maintenance (often defined as occurring during specified maintenance windows, with advance notice), force-majeure events, third-party failures, problems on your end (your network, your browser), security incidents that require taking the service down, and any downtime "caused by factors beyond the vendor’s reasonable control." Each exclusion reduces what the percentage actually guarantees.

Scheduled maintenance is the biggest one to read closely. If a vendor reserves the right to take the service down for up to four hours a week for maintenance, the "99.9% uptime guarantee" only applies outside those weekly windows. So the real uptime guarantee, including planned downtime, could be much lower than the headline number suggests. The fix is not necessarily to oppose scheduled maintenance, vendors need it, but to ensure the windows are reasonable (off-hours for your business), advance notice is required, and total maintenance time per month is capped.

What do you get when a SaaS vendor misses its SLA?

When a vendor misses its SLA, what do you get? Almost always: a service credit applied to your next invoice. The credit is typically calculated as a percentage of your monthly fee based on how much the vendor missed the target. A common structure is:

  • Uptime between 99.0% and 99.9% in a month: 10% credit on that month’s fees.
  • Uptime between 95.0% and 99.0%: 25% credit.
  • Uptime below 95.0%: 50% to 100% credit.

Why don't service credits match your real losses?

Service credits look like a remedy, but for most customers they bear almost no relationship to the actual cost of downtime. If your business depends on a SaaS tool and the vendor has a four-hour outage during a peak day, your real loss may be tens of thousands of dollars in missed sales, employee productivity, and customer-trust damage. The SLA "remedy" is a 10 percent credit on a $200 monthly fee, twenty dollars. The economics are designed to let the vendor keep their commercial promises without exposing themselves to actual damages.

This is not unreasonable for routine outages, where small credits acknowledge a fall short of the standard without bankrupting the vendor. But for prolonged or severe outages, the gap between the credit and the actual harm can be vast. Negotiating higher credit percentages, lower thresholds, and termination rights for sustained failures is how you align the remedy with the harm. Treating service credits as a real safety net for material outages is generally a mistake.

Can a service credit legally be your only remedy?

Many SLAs say service credits are your "sole and exclusive remedy" for downtime, and contract law generally lets businesses agree to that. The sales rules in UCC § 2-719(1)(b) say an agreed remedy is optional unless the contract expressly makes it exclusive, which is why vendors use that exact phrase. Courts disagree on whether Article 2 governs SaaS at all, but they often borrow its principles.

There is a safety valve. Under § 2-719(2), if an exclusive or limited remedy fails of its essential purpose, the ordinary remedies become available again. Courts have mostly applied this to repair-or-replace promises that the seller never fulfils, and whether chronic outages paid off with token credits would qualify depends heavily on the facts and the state. It is not something to rely on.

The exclusive-remedy wording also usually sits alongside a general liability cap and an exclusion of lost profits, so even a successful argument may not recover much. That is why the contract terms covered here, higher credits, narrower exclusions and above all a right to terminate for repeated failures, are the protections worth negotiating. A termination right does not depend on persuading a judge that the remedy failed.

You have to claim the credit

A detail that catches many customers off guard: most SLAs require you to formally request the credit, in writing, within a defined window after the outage (often 30 days). Vendors do not proactively apply credits, they wait for you to notice the outage, document it, and submit a claim citing the specific incident, your downtime evidence, and the SLA terms. If you do not submit a claim within the window, the credit is forfeited. So even an enforceable SLA only pays out if you operationally track outages and follow the claim process. Many customers never bother for small incidents, which is what the structure is designed for.

Termination rights for chronic failure

The most valuable SLA term, often missed, is a termination right for chronic or severe failure. A clause letting you terminate the contract without penalty if the vendor misses the SLA in multiple consecutive months, or has a single sustained outage beyond a threshold (often 24 hours), turns the SLA into an actual exit valve. Without it, even repeated SLA failures only generate credits, you are stuck in the contract until renewal, paying for a service that has been demonstrably unreliable. With a termination right, you have a meaningful response to systemic problems.

A typical clause might say: "Customer may terminate this Agreement without penalty if the Service falls below the SLA target in three consecutive months, or if any single outage exceeds 24 continuous hours. In addition to any termination, Customer will receive a refund of any prepaid, unused fees." That kind of clause aligns the vendor’s incentives with actually meeting the SLA, because chronic failure costs them the customer and the contract, not just a small credit.

Status pages and incident transparency

A reliable vendor will publish a status page tracking incidents and uptime, with public history you can verify against the SLA. Before you sign, look at the vendor’s status page and see how it tracks against their advertised SLA over the prior 12 months. Frequent or unexplained gaps are signals. Some vendors require their own internal incident records to be the authoritative source for SLA calculations, which is awkward, if they decide an incident "did not count," you have little recourse. A fair SLA either references a public status page or commits to provide reasonable documentation of any outage on request.

Match the SLA to your actual dependency

How much SLA detail matters depends on how critical the service is to you. For a SaaS tool that supports a non-essential internal process, a standard 99.9% SLA with basic credits is usually fine. For a service that powers your customer-facing product, your e-commerce checkout, or any function where downtime directly costs money, the SLA suddenly matters a lot and deserves serious negotiation. Match the energy you spend on the SLA to the actual cost of an outage in your business. Over-negotiating an SLA you will never need to invoke is a poor use of your leverage; under-negotiating one where downtime would be catastrophic is the more expensive mistake.

How is SaaS uptime measured, and who decides?

A subtle but consequential point: who measures uptime, and how? Some SLAs specify external monitoring against defined endpoints. Many simply say uptime is measured by "the vendor’s monitoring systems," which means the vendor is judge and jury of whether they hit their own target. That sounds suspicious, and it can be, but in practice most vendors track uptime accurately because their own engineering teams need real numbers. The risk is when something is borderline, or when the vendor measures only at a coarse level (the API responding "OK") while the actual product is degraded enough to be unusable for you.

The fix is to define "downtime" in a way that matches your real experience. Downtime should mean the service is unavailable or materially degraded for your use, not just a narrow technical check. Asking the vendor to use an external monitoring source, or at least to provide their incident data on request, brings the measurement closer to ground truth. Without that, the SLA percentage is essentially what the vendor tells you it was, and disputes about borderline outages are very hard to win.

What to negotiate

When the SLA matters, the targets to push on are:

  • A higher uptime commitment (99.95% or 99.99%) if appropriate to the service’s criticality.
  • Narrower exclusions, cap scheduled maintenance hours, require advance notice, restrict the "beyond reasonable control" language.
  • Higher credit percentages and lower thresholds, so even moderate misses pay meaningful credits.
  • Automatic credit application rather than a claim requirement, where possible.
  • A termination right for repeated or sustained failures, with refund of prepaid fees.
  • Public-status-page reference as the source for outage tracking.
  • A separate, higher SLA for critical-priority support response times (not just uptime).

The bottom line

A 99.9% SLA permits roughly 8.7 hours of downtime per year, with remedies that almost always undervalue the cost of an outage, unless you negotiate them. Read the uptime number alongside the exclusions, the credit schedule, the claim process, and any termination rights. For mission-critical services, push for tighter targets, narrower exclusions, and a real exit if the vendor cannot deliver. If you want a fast read on what your vendor’s SLA actually promises and what it really pays you when they miss, ClauseAudit reviews the agreement in about a minute and flags every exclusion, credit threshold, and missing termination right, so you can negotiate the parts of the SLA that actually matter to your business.

Check what your SLA actually guarantees

Upload your SaaS contract and we will flag uptime commitments, exclusions and service credits, plus every other risky clause, in plain English, tuned to your state, with a downloadable report and redline.

Free templates for this

Editable, plain-English starting points with the fair versions of these clauses.

Frequently asked questions

What does 99.9% uptime actually mean?

About 8 to 9 hours of allowed downtime per year. 99.5% is roughly 44 hours, and 99% is more than three and a half days, so small percentage differences translate into very different real-world reliability.

Do SLA service credits cover my losses?

Rarely. Credits are usually a small percentage of your fees, not compensation for your actual business impact, and you typically have to file a claim within a short window to receive them at all.

What should I negotiate in an SLA?

A meaningful uptime target for your dependency, narrow exclusions, credits that are automatic rather than claim-only where possible, transparent measurement, and a right to terminate if the vendor repeatedly misses the SLA.

Does scheduled maintenance count against a SaaS uptime SLA?

Usually not. Most SLAs exclude scheduled maintenance from the uptime calculation, so a 99.9% promise may apply only outside the maintenance windows. If a vendor can take the service down for hours each week, real availability can be far lower than the headline. Ask for off-hours windows, advance notice, and a monthly cap on total maintenance time.

Do I have to claim SLA service credits myself?

In most SLAs, yes. Vendors rarely apply credits automatically. You typically must submit a written claim within a set window after the outage, often 30 days, citing the incident, your downtime evidence and the SLA terms. Miss the window and the credit is forfeited, so track outages as they happen or negotiate automatic credit application.

Can I terminate a SaaS contract for repeated outages?

Only if the contract gives you that right, so negotiate it. A strong clause lets you terminate without penalty if the vendor misses the SLA in several consecutive months or has a single outage beyond a threshold, often 24 hours, with a refund of prepaid unused fees. Without it, repeated failures only earn small credits until renewal.

What is an SLA in a SaaS agreement?

An SLA, or service level agreement, is the part of a SaaS agreement where the vendor commits to measurable performance, most often a monthly uptime percentage such as 99.9%. It also sets support response times, exclusions like scheduled maintenance, and the remedy if the vendor misses the target, usually service credits you must claim. Credits rarely match your real losses.

Related guides

This guide is general information from ClauseAudit, not legal advice. Laws vary by state and change, consult a qualified attorney for your situation. Published 2026-05-01; last reviewed 2026-10-09.