1. Response time and repair time — defined separately
This is the most common gap in weak agreements: a single general "resolution time" that means nothing concrete. Response and repair are two different commitments. Response is acknowledging the ticket and starting work — proof that the service has moved at all. Repair is actually removing the failure and restoring operation. The agreement must define both deadlines separately, because a fast response with no commitment to a repair deadline is a false sense of security — the service "responded" with a phone call, and the system is still down.
2. Failure categories and priorities
A failure that stops production and a burnt-out corridor bulb can't have the same response time — because either the first will be handled too slowly, or you'll overpay for the second. A good agreement splits failures into categories by their impact on the facility's operation and assigns each its own, adequate response and repair times. Without that split, you either pay for priority handling of trivia, or wait for a critical failure to be addressed as long as for a fluorescent tube.
3. Hours covered by the agreement
Service 24 hours a day, 7 days a week, or only during business hours — this is one of the few things that translates directly into the subscription price, and so must be stated unambiguously. A failure rarely picks a convenient time. An agreement silent on this, or hiding the limitation in the fine print, leaves the client with a dead system on Friday evening and service available only on Monday morning.
4. Scope of systems covered by the agreement
This is the clause whose absence most painfully hits facilities with mixed infrastructure — and most of them are. Does the agreement cover only installations built by this service provider, or also the other contractors' systems present in the facility? Without a clear answer, the client discovers the limits of coverage only at a failure, when told that this particular system "wasn't in scope." A facility with five systems from three contractors and one service agreement covering only some of them has gaps exactly where it least expects them.
5. Reporting after every intervention
Every intervention should end with a report: cause of the failure, work performed, recommendations for the future. Written into the agreement as an obligation, not left to the service provider's goodwill. This isn't bureaucracy — it's how a facility's history is built, one that over time itself suggests what to replace before it fails. Service without reporting is a series of disconnected repairs with no memory; service with reporting is growing knowledge of how the facility ages and where its weak points lie.
6. Penalties for exceeding the SLA
A declaration of response time with no consequence for exceeding it is a wish, not a commitment. A penalty clause — even a symbolic one — changes the character of the whole agreement: deadlines stop being a promise and become a commitment whose breach costs the contractor something. This isn't a clause aimed at the service; it's a clause that levels the interests of both sides and makes the times declared in the agreement taken seriously, because they carry a price.
7. Escalation path
The last, most often omitted clause, and often the most important: what happens when the first line of service doesn't respond within the agreed time. Who takes over the case, by when, on what number. Without an escalation path, the client — in a situation where the standard procedure fails — is left alone with a dead system and no answer, because they don't know whom to turn to when an ordinary ticket didn't work. A well-designed escalation is the guarantee that a failure won't stall in a dead end just because one person didn't pick up the phone.
A document measured on failure day, not signing day
A good SLA agreement isn't a matter of page count or the density of legal language, but of the presence of these seven clauses. The rest is formality that doesn't matter on failure day. It's also worth noting how the contractor even measures SLA delivery — a declared response time is worth as much as the way it's counted. A service in which response and repair times are logged in a system, and the status of every ticket is available to the client in real time, turns the SLA from a promise into a document verifiable at any moment — rather than only at month-end, on the basis of assurances.
An SLA agreement is judged not on the day it's signed, when everything works, but on the day of the first serious failure, when nothing does. These seven clauses decide which side of that test you'll be on.