Which SaaS vendors actually pay SLA credits?

Plenty of vendors publish an SLA. Fewer make it easy to collect on it — the credit percentages, the deadlines, and even whether there's a published tariff at all vary widely. This is a side-by-side of the vendors Ontracko monitors that carry a credit remedy, drawn from their published SLAs.

The short answer

The major cloud providers — AWS, Azure, and Google Cloud — all pay tiered SLA credits, as do Atlassian, Twilio, and GitHub (Enterprise). Cloudflare pays a formula-based credit under its Enterprise SLA — and runs separate, conventional tier tables for Workers, R2 and Queues. Salesforce, Slack, and Zoom publish no tariff at all: whatever they owe is set in your own agreement. Across all of them the credit is a percentage of the affected month's fee, and most require you to file within 30 days — Atlassian gives you 15, and Cloudflare wants notice within 5 business days — so the deadline, not the vendor, is usually what stands between you and the money.

But the more useful answer is the one nobody advertises: a published uptime number does not mean a credit exists. We read the actual contract for every vendor we monitor, and seven of them — Datadog, Twilio Segment, New Relic, Asana, Postman, Amplitude and CircleCI — carry no service-credit remedy at all, even though several were publicly associated with a 99.9% SLA. An eighth, LaunchDarkly, has a real credit that almost never triggers. Skip to that list if that is what you came for.

The comparison

Vendors that publish an uptime number but pay no credit

This is the part of the market that costs companies the most time, because the number is easy to find and the remedy is not. Each of the vendors below was checked against its own published agreement — not a marketing page, not a third-party summary — and in each case there is no service credit to claim.

Two of these deserve a warning. Amplitude and CircleCI were both widely described as having a 99.9% SLA — we carried that figure ourselves until we read the agreements and found no uptime commitment in either. If a vendor's uptime number only ever appears in blog posts and comparison tables and never in its own legal document, treat it as marketing, not a contractual commitment.

LaunchDarkly is the instructive edge case. Its Enterprise Support tier — the common one — has no credit at all; the sole remedy is a right to terminate. Only Premium Support carries money (10%/25%, capped at 50%), and even there the credit is gated: it pays only after two consecutive months below the commitment, or a single month below 75% availability. So an ordinary bad month pays nothing. The trap is the deadline: the SLA claim must be filed within five days of the incident, long before anyone can know whether the following month will also miss. If you wait to find out, the right is gone. File the first month anyway, on time, even though that month alone owes you nothing.

That is the general rule worth taking away: file to preserve the right, even when this month's arithmetic says zero. A claim never lodged cannot be revived later.

Reading the table

The cloud providers are the most generous — and the most conditional. AWS, Azure, and Google Cloud can credit up to 100% of a service's monthly fee for a severe outage, but the higher tiers only trigger when uptime falls far below target, which is rare. Most real-world claims land in the first tier (10%). The windows are not comparable as bare day counts, because they are anchored differently: Azure allows 60 days from the incident, most Google Cloud services 30 days from the point you become eligible (Compute Engine 60), while AWS runs to the end of the second billing cycle after the incident — which can be nearly 90 days for an outage early in a month, and barely 60 for one at month-end.

The plan matters more than the logo. SLAs almost always apply to paid or enterprise tiers only — Atlassian credits apply to Premium and above, Cloudflare's Enterprise SLA is the document we have read (its self-serve plans are governed by a separate agreement we have not), and GitHub's applies to Enterprise Cloud. On a free or entry plan, there may be no credit remedy at all.

Some vendors publish neither a number nor a tariff. Twilio publishes its 99.95% commitment and sets the final credit percentage in your order form — so there the SLA page tells you the target and the deadline, and only the credit lives in your contract. Salesforce and Zoom go further: neither publishes an availability percentage at all, and both settle any commitment *and* any credit through your Master Subscription Agreement or order form. Slack is the furthest case of the three, because it does publish a document called a Service Level Agreement and that document promises no number: it commits Salesforce (which owns Slack) to "commercially reasonable efforts" to keep the service available, and defines no credit and no deadline. For all three, your own signed agreement is the only primary source — read it before quoting yourself a figure.

Cloudflare is the outlier. Its Enterprise SLA commits to 100% uptime and computes the credit from a formula — the outage period, the affected-customer ratio and scheduled availability, scaled by your success package — rather than a tier table. Cloudflare publishes that formula only as an image, so we do not reproduce the multiplier. Its Workers, R2 and Queues SLAs are ordinary tier tables and control over the Enterprise terms for those services.

You can open any vendor above for its exact tier schedule and a live credit calculator, or see who's actually been going down lately on the reliability rankings.

How to claim a credit from any of these vendors

1. Confirm the outage from the vendor's public status page — capture the incident reference ID and the exact start and end times.
2. Measure your affected service's monthly uptime and match it to the vendor's credit tier.
3. Collect the required evidence: your account/organization ID, the affected service or region, and the incident timestamps.
4. Open a billing or support case and request an SLA "service credit" under the availability clause, attaching your calculation and the incident reference.
5. File before the window closes — 30 days for most vendors, 60 for AWS and Azure — because the deadline runs from the incident, not from when you noticed.

Frequently asked questions

Which SaaS vendors advertise an SLA but pay no credit?

Reading the vendors' own agreements rather than their marketing pages, seven pay no service credit: Datadog, Twilio Segment and New Relic offer a termination right instead of a credit; Asana publishes a quarterly 99.9% target with no credit table and no claim process; Postman offers termination only, and only after three consecutive missed months; and Amplitude and CircleCI publish no availability SLA at all, despite both being widely listed as 99.9% vendors. LaunchDarkly technically has a credit, but only on Premium Support and only after two consecutive missed months.

Does a published uptime percentage mean I am owed money when it is missed?

No. An uptime commitment and a credit remedy are two different clauses, and plenty of agreements contain the first without the second. Before you spend time building a claim, confirm the agreement actually contains a service-credit schedule and a claim process — and check which support tier or plan it applies to, because credits are very often restricted to the enterprise tier.

Which cloud provider pays the highest SLA credit?

AWS and Azure can both credit up to 100% of a service's monthly fee for the most severe outages, versus Google Cloud's 50% cap. But the top tiers require uptime to fall far below target; most claims resolve at the entry tier of around 10%.

Do all SaaS vendors pay SLA credits?

No. A credit remedy exists only where the vendor publishes (or contractually agrees) an SLA with a credit clause, and it almost always applies to paid or enterprise plans. Free and entry-tier plans frequently have no credit remedy.

Which vendor has the shortest claim window?

Among the major clouds, Google Cloud's 30 days is the shortest count, but the anchor matters more than the count: Google's 30 days and Azure's 60 both run from the incident, while AWS's window runs to the end of the second billing cycle after it, so an AWS outage early in a month can carry nearly 90 days. Among SaaS vendors Twilio and GitHub use a 30-day deadline, Atlassian is 15 days, and Cloudflare requires notice within 5 business days of the incident. Slack, Zoom and Salesforce set no published deadline at all, because they publish no credit for one to run against — check the notice period in your own agreement.

Does Cloudflare pay SLA credits?

Yes. Its Enterprise SLA carries a 100% uptime commitment and a formula-based credit rather than a tiered table, and requires notice of intent within 5 business days of the incident. Workers (99.99%), R2 and Queues (99.9%) are governed by their own SLAs with 10/25/50% tier tables, and you may not claim under one of those and the Enterprise SLA for the same incident.

Methodology & caveats

Every figure here is transcribed from the named vendor's published SLA into Ontracko's monitored profiles, current as of the updated date above — and where a vendor publishes no figure, this page says so rather than printing one. That distinction is the correction of a real error: until 2026-09-06 the Slack and Zoom rows above carried a committed percentage and a credit ceiling that appear in no Slack or Zoom document, because the profiles behind them cited the wrong source — Slack's end-user terms of service rather than its Service Level Agreement. Credit percentages, caps, and deadlines can change and often depend on plan and negotiated terms; for vendors whose credits are set in an MSA or order form (Salesforce, Slack, Zoom, and Twilio's percentage), confirm your own agreement before filing.

*Ontracko monitors SaaS & cloud vendors' public status feeds and recovers the SLA credits when they miss. Free — 8% only on recovered credits. See live reliability rankings or start with what is an SLA credit.*