Service-Level Agreement (SLA)

A Service-Level Agreement (SLA) is a formal, usually contractual, commitment a vendor makes to a customer defining a measurable service standard — most commonly uptime/availability, but also support response times, data-processing latency, or resolution times for reported bugs — along with specified remedies (typically service credits, i.e., a partial refund of subscription fees) if that standard is breached. SLAs are a standard, often deal-breaking requirement in mid-market and enterprise SaaS sales: a company won't run mission-critical operations on a vendor's product without a contractual guarantee, backed by financial consequences, that the vendor takes availability seriously. SLA uptime commitments are conventionally expressed as a percentage of "nines" — 99.9% ("three nines") allows roughly 43 minutes of downtime per month, while 99.99% ("four nines") allows only about 4.3 minutes per month — and the difference between these tiers requires meaningfully different infrastructure investment (multi-region failover, zero-downtime deploys, redundant everything), which is why higher SLA tiers are often reserved for a vendor's higher-priced enterprise plan. Crucially, an SLA is a legal/contractual commitment with financial teeth, distinct from uptime itself (the actual measured metric) and from a status page (the public-facing communication of that metric) — a vendor can have excellent actual uptime with no formal SLA at all (common for early-stage startups not yet selling to enterprise), or a formal SLA whose penalty (a small service credit) is trivial compared to the customer's real business impact from an outage. Concrete worked example: an enterprise contract includes an SLA clause guaranteeing 99.95% monthly uptime (about 22 minutes of allowed downtime), with a remedy schedule of a 10% service credit for uptime between 99.0–99.95%, and a 25% credit below 99.0%. In a month where the vendor suffers a 90-minute outage (99.79% actual uptime — a breach of the 99.95% commitment but not the 99.0% threshold), the customer is contractually owed a 10% credit on that month's invoice, which the vendor's billing system must be able to calculate and apply automatically or via a documented claims process. SLAs increasingly extend beyond raw uptime for AI-powered SaaS products specifically, adding commitments around model response latency, API error rates, or even data-processing accuracy thresholds — reflecting that "is the service reachable" is no longer the only dimension of service quality enterprise buyers care about once inference latency and output quality become part of the product's core value proposition. Beyond service credits, some enterprise SLAs also include a termination-for-convenience clause triggered by repeated or severe breaches, giving the customer a contractual escape hatch to exit the agreement early without penalty if the vendor's reliability consistently falls short — a stronger remedy than a service credit alone for a customer whose business genuinely cannot tolerate ongoing instability from a mission-critical vendor.