Commercial Transactions

SLAs and Service Credits: What the Uptime Promise Actually Gets You

Every software vendor promises high availability. The service level agreement is where that promise gets measured, limited, and priced.

An SLA that says "99.9% uptime" sounds like a commitment. Whether it actually protects the customer depends on how uptime is measured, what is excluded, what the customer gets when the vendor misses, and whether that remedy is the only one available. Those details are usually in the fine print, and they matter more than the headline number.

This article covers how SLAs are built, where the negotiation happens, and how the SLA interacts with the rest of the contract.

What the Percentage Means

Availability commitments are usually expressed as a percentage of time in a measurement period, most often a calendar month. The differences between common tiers look small on paper and are not small in practice:

  • 99.5% allows roughly 3.6 hours of downtime in a 30-day month.
  • 99.9% allows roughly 43 minutes.
  • 99.99% allows roughly 4 minutes.

The right tier depends on what the service does. A reporting tool used once a week can tolerate more than a payment processing system. Customers generally benefit from negotiating for the tier that matches the business impact of an outage. Vendors should commit only to what their infrastructure and upstream providers can actually support.

Measurement Is the Real Negotiation

Two SLAs with the same percentage can provide very different protection depending on how "available" is defined.

What counts as down. Is the service unavailable only when nothing loads at all? Or does severe degradation count, such as error rates above a threshold or core features failing? Customers should look for a definition that covers the functions they actually rely on.

Who measures. Most SLAs rely on the vendor's own monitoring. That is often reasonable, but customers can ask for access to status data, or for the right to submit their own logs as evidence in a credit claim.

Granularity and scope. Check how often availability is sampled (every minute or every five), whether a partial or regional outage counts, which modules and APIs are covered, and whether beta or free-tier features are carved out. An outage that only hits one region or one critical API can be invisible to a coarse measurement.

Exclusions. Every SLA excludes something. Standard exclusions include scheduled and emergency maintenance, customer-caused issues, force majeure events, and problems with the customer's own systems or internet connection. The questions to ask are whether scheduled maintenance has a notice requirement and a monthly cap, whether it must occur during off-peak hours, and whether failures of the vendor's own hosting provider are excluded. A vendor that excludes outages caused by its cloud provider may have excluded one of the most significant categories of outage risk. A common compromise excludes only region-wide provider failures that go beyond the redundancy the vendor has committed to maintain.

Service Credits

When the vendor misses the commitment, the usual remedy is a service credit: a percentage of the monthly fee applied against a future invoice. An illustrative schedule might provide a 10% credit below 99.9%, 25% below 99%, and 50% below 95%. Actual tiers vary widely.

Three details determine whether credits are worth anything:

  • Claim procedures. Many SLAs require the customer to request a credit in writing within a short window, sometimes 30 days, with supporting detail. Miss the window and the credit is gone. Customers often calendar the deadline or negotiate automatic credits. Some SLAs also make credits available only if the customer is current on payment.
  • Caps. Credits are almost always capped, often at a percentage of the monthly fee for the affected service. Caps are often well below 100% of the monthly fee, and even a 100% cap limits annual exposure to about one-twelfth of the contract value.
  • Form. Credits applied to future invoices have little value if the customer is leaving. Customers can ask that unused credits be paid as a refund at termination.

Service credits are rarely sized to compensate for the actual business loss caused by an outage. That is by design. They are a price adjustment for a service that did not perform as promised. Meaningful protection against a failing vendor usually requires an exit right, along with the ability to export data and get transition help on the way out.

Sole and Exclusive Remedy

This is the clause that deserves the most attention. Many SLAs state that service credits are the customer's "sole and exclusive remedy" for any failure to meet the service levels.

From the vendor's side, that language is important. It keeps a downtime event from becoming a breach of contract claim for lost profits, and it makes SLA exposure predictable.

From the customer's side, an exclusive remedy clause without an exit right can mean the customer is stuck paying for a service that keeps failing, collecting small credits along the way. The common compromise is a chronic failure termination right: if the vendor misses the SLA in, for example, three months out of any rolling six-month period, the customer can terminate without penalty and receive a refund of prepaid, unused fees.

The contract should also say whether the exclusive remedy language bars a termination for material breach under the main agreement. How a court reads these provisions depends on the drafting and the governing law, so the interaction is worth spelling out rather than leaving to argument.

The exclusive remedy language should also be read alongside the contract's limitation of liability provisions. An outage caused by a security incident such as ransomware, for example, may be both an SLA failure and a breach of security obligations. The contract should make clear that the exclusive remedy clause applies only to availability, not to separate obligations around security, confidentiality, or data protection.

Support Response Times

Many SLAs also set support commitments, usually tiered by severity. A critical issue might require a response within one hour, a minor issue within one business day. Two points to check:

  • Response is not resolution. A one-hour response time means someone acknowledges the ticket. It does not mean the problem is fixed. Customers who need resolution commitments should ask for them separately, though vendors often resist hard resolution deadlines.
  • Who sets severity. If the vendor alone decides whether an issue is critical, the response-time commitments are only as strong as the vendor's classification. A shared or objective definition works better.

What an Availability SLA Usually Does Not Cover

An uptime commitment says nothing about whether the customer's data survives an incident. Backups, data loss, and recovery time and recovery point objectives typically need their own commitments, often in the security exhibit or DPA. Customers that care about what happens after a major outage can also ask for a written incident report or root-cause analysis within a set period.

The other quiet issue is change. Many SLAs live at a URL and apply "as updated from time to time." Customers can ask that updates not reduce service levels during the current term. For many customers, this is the single most important edit to a vendor's SLA.

Practical Takeaways

For customers: focus less on the headline percentage and more on the definition of downtime, the exclusions, the claim process, and the exit right. A 99.9% SLA with a chronic failure termination right is often worth more than a 99.99% SLA without one.

For vendors: commit to what your architecture can deliver, keep the exclusive remedy language clear, and expect sophisticated customers to ask for a chronic failure exit. Offering one proactively, with reasonable thresholds, can shorten negotiations.

The SLA is one piece of a larger contract stack. For how it fits with the MSA, order form, and DPA, see our overview of the commercial contract stack.

This article is general information as of November 2026. It is not legal advice and does not create an attorney-client relationship. Contract outcomes depend on the specific terms, facts, and governing law.

Related reading

How we can help

Let's Get Started