Service level agreement (SLA)
A service level agreement is the written part of a vendor contract that defines what work is covered, how quickly the vendor will respond and resolve issues, during which hours, and what remedy applies when they miss those commitments.
In practice
An SLA is the difference between a promise and a term you can hold someone to. Most website maintenance contracts contain something described as an SLA that specifies none of the four things that matter: what counts as an incident, how fast the vendor answers, how fast they fix, and what you get if they do neither. A document that says the vendor will respond promptly has defined nothing.
The distinction that causes the most damage is response time versus resolution time. Response time is how long before a human acknowledges your ticket. Resolution time is how long before the problem is fixed. A four-hour response commitment with no resolution commitment is compatible with a site being down for a week, and vendors know this even when clients do not. Ask for both, and accept that resolution times are often stated by severity rather than as one number, which is reasonable.
Severity tiers are worth insisting on for the same reason. A checkout or quote form that is down is not the same class of problem as a typo on an interior page, and an agreement that treats them identically will either overprice the typo or underserve the outage. Two or three tiers with an example of each is enough. More than that becomes a document nobody consults.
Coverage hours are the quiet clause. A commitment measured in business hours means a Friday evening failure starts its clock on Monday morning, which for a B2B company whose buyers research in the evening can mean a full weekend of lost quote requests before anyone is contractually late. That may be an acceptable trade at your price point. It should be a decision rather than a discovery.
Uptime percentages deserve translation before you accept one. Over a 30 day month, 99.9 percent allows about 43 minutes of downtime, 99 percent allows about seven hours, and 95 percent allows about 36 hours. Those are very different products described in similar language, and the arithmetic is worth doing at the table.
Here is the test. Ask your vendor what happens at hour five of an outage that started at 6pm on a Friday: who is awake, what have they committed to, and what do you receive if nothing was done until Monday. A vendor with a real SLA answers in specifics because the answer is written down. A vendor without one describes their general attentiveness. Both replies are useful; only one is enforceable.
Common questions
What should a website SLA include?
A definition of what counts as an incident, severity tiers with examples, separate response and resolution commitments per tier, the hours during which those clocks run, an escalation path with names, and the remedy when a commitment is missed. Anything less is a statement of intent rather than an agreement.
What is the difference between response time and resolution time?
Response time is how long before someone acknowledges the problem. Resolution time is how long before it is fixed. A contract that commits only to response time is compatible with a long outage, which is why the two should always be stated separately.
What does 99.9 percent uptime actually allow?
About 43 minutes of downtime in a 30 day month. For comparison, 99 percent allows roughly seven hours and 95 percent allows roughly 36 hours. Doing that arithmetic before signing turns a percentage into something you can judge against what an outage costs you.
Do we need an SLA for a small B2B website?
You need the parts of one. Even a short agreement is worth having if it names response and resolution targets, coverage hours and an escalation contact. The value is not contractual enforcement so much as having agreed what normal looks like before something breaks and both sides are improvising.