1. What an SLA is and what it is not
SLA is the abbreviation of Service Level Agreement, the agreement that sets the level of service expected from a provider. In practice it describes what is guaranteed, with which response and restoration times, which exceptions apply and how the result is measured. It is not a general commitment to do your best. A serious SLA is verifiable: every commitment has an indicator and a means of checking it.
Many companies confuse the SLA with the price list or with the opening hours of the support desk. These are related but different. The hours tell you when you can call; the SLA tells you what happens after the call and within what timeframes. Support open many hours a day with a vague SLA can be less useful than support with more limited cover but clear rules.
2. Start from the impact on work, not from the technology
The most effective way to define a service level is to ask what happens when a system does not work. A problem with email at a firm of a few people does not carry the same weight as a block on the management system during month-end closing, or a production line that has stopped. The technology involved matters less than the effect on work and on people.
For each critical process, write down three things: who depends on it, how long it can stay down without serious harm, and at which times it is really needed. This exercise usually reveals that not everything is equally urgent. Setting everything at the highest level is expensive and does not improve continuity, because the provider ends up with no clear priorities.
3. Classify requests by priority
Good classification allows the provider to know which problem to tackle first, without discussion every time. Priority levels are usually linked to impact and urgency: a total outage of a critical service is different from a problem that has a workaround. Define the criteria in plain words, with concrete examples from your systems. A simple table, shared with the provider, avoids arguing about the priority of each specific fault.
| Priority | Generic example | What to expect | Who decides the priority |
|---|---|---|---|
| Critical | An essential service is down for all users | Immediate pick-up and frequent updates | The company contact, confirmed by the provider |
| High | A group of users cannot work, with only a slow alternative | Rapid action within the agreed hours | The company contact |
| Medium | A single problem with a temporary workaround available | Planned action by the next working day | The provider, in agreement with whoever reported it |
| Low | A change or information request | Ordinary handling according to the work queue | The provider, under the terms of the contract |
4. Choose the right indicators
An SLA without indicators is a statement of intent. The most commonly used indicators measure the time between opening and pick-up, the time between pick-up and resolution, the availability of critical services and the number of reopened requests. Do not ask for everything: choose a few indicators linked to the processes you have mapped. Every extra indicator is another control activity, and risks going unused.
For each indicator define how it is calculated, which tracking system it comes from and what exceptions it allows. An indicator that cannot be verified by both parties creates pointless arguments at the end of the month. Check with the provider that its request management system records the necessary dates, otherwise the measure does not exist. Ask for a sample report before signing, because proof of performance lies in the data, not in promises.
5. Define exclusions and responsibilities
Exclusions are the most important part of an SLA, and often the least read. They specify what falls outside the commitment: work on systems supplied by third parties, faults caused by changes made by unauthorised staff, events outside the provider's network. They are not a loophole if they are written precisely; they become one when they are generic. Each exclusion must be read together with the service description, to understand whether it empties of meaning a commitment already written.
Also clarify shared responsibilities. If the service depends on an internet line provider or on software run by another party, your SLA must say who coordinates resolution and who communicates with users. Without this clarification, every problem turns into a ping-pong between two suppliers, and the restoration time grows for everyone.
6. The step-by-step path
The design process requires more discussion than technical skill. Following a precise order avoids defining commitments you have no way of checking. Here is the sequence that usually works best. You can carry it out with internal staff and the provider in a few short meetings, provided each step produces a written document. If the provider suggests skipping a step, ask why: that is often where the next problem begins.
- List the services covered by the contract and the processes that depend on each
- Assign each process an impact and an urgency, with real examples
- Define the priority classes and the criteria for assigning them
- Choose a small number of indicators and how they are calculated
- Write the exclusions and shared responsibilities in precise terms
- Agree the escalation procedures and communication channels
- Set a periodic review of the SLA and a criterion for amending it
7. The most frequent mistakes
Some mistakes recur in almost every badly written SLA. The first is promising resolution times for problems that depend on third parties, such as a line or software provider. The second is copying a generic template without adapting it to the company's processes. The third is having no indicator for what is promised: the commitment remains on paper.
When you find one of these errors in the draft, do not simply correct it. Ask the provider to explain how it will handle the matter in practice: the answer shows whether the problem has been understood or just copied. A diligent provider welcomes the request, because it reduces the risk of future disputes for both sides. It is also more comfortable for the provider to work with defined rules, since it knows exactly what it must do and how it will be judged.
- Commitments at the same level of urgency for everything
- Generic exclusions such as causes not attributable to the provider
- Indicators that nobody can measure or that the provider does not record
- No review of the SLA after major changes in the company
- Penalties or credits with no clear calculation criterion
8. How to check that the SLA works
An SLA is proven over time, not at the moment of signing. It needs regular review of the data, at the frequency set by the contract, in which the indicators are compared with users' actual experience. When the numbers are good but people are complaining, the problem is probably in the indicator, not in the people. In that case it is better to correct the measure rather than the judgement about the people involved.
Bring to the review the indicator data, the most significant incidents with their causes and the recurring requests that should not exist. This last item is often the most useful, because it shows where to intervene preventively. If the provider cannot supply this data, that is a signal to take seriously before renewal. Timeliness in providing data is also part of the service and should be assessed like everything else.
9. How to start
The first feasible step this week is to choose a single critical process and describe it on one page: who uses it, at what times, what happens if it stops for a whole day. No new tools are needed. With this sheet you can already see which commitments to request urgently and which can remain lighter. Focus on one activity at a time, because an analysis that is too broad tends to remain incomplete.
Then compare the page with the contract you have or with the draft you have been offered. Note every commitment that has no indicator, every generic exclusion and every unclear responsibility. Take the list to the provider: it is the quickest way to turn a commercial document into a service level you can check. The result is a short document that both parties can read and sign without ambiguity.
Frequently Asked Questions
What is the difference between an SLA and support opening hours?
Support hours tell you when you can request assistance. The SLA tells you what happens after the request: how quickly it is picked up, which priorities apply, how it is resolved and how it is measured. Support available across many time slots may have a weak SLA, while a service with limited hours may have very clear commitments. Assess both together.
Must the SLA be the same for all services?
No. A service critical to production requires tighter commitments than a rarely used office workstation. Applying the same level to everything costs more and does not improve continuity. Differentiate services according to the impact they have on work, and concentrate resources where downtime causes real damage.
Can I claim penalties if the SLA is not met?
You can ask for credits or other forms of compensation, but the value and mechanism depend on the contract. More important than the penalty is that the mechanism is clear: which indicators trigger it, how they are calculated and who verifies them. A vague penalty brings no real advantage and can create disputes. If one is provided for, ask that the calculation is set out in writing in the contract.
How often should the SLA be reviewed?
A periodic review set out in the contract is usually advisable, plus an extraordinary review after major changes: new systems, new sites, new activities or staff changes. During the review check whether the indicators are still useful and whether the exclusions cover the situations that have actually occurred. An SLA unchanged for years is often an outdated SLA.
Who should define the indicators, the company or the provider?
Ideally both parties together. The company knows which processes matter and what effect a stoppage has; the provider knows what data it can produce with its own tools. An indicator the provider cannot measure is useless, and one the company does not understand will not be used. Define it in a working session and record it in the contract with the calculation formula.
Does an SLA alone guarantee business continuity?
No. The SLA describes the provider's commitments, but continuity also depends on technical choices, internal processes and emergency preparedness. The SLA is a governance tool: it helps you take decisions and verify the service, but it does not replace a written and tested business continuity plan.
How do I handle an SLA the provider cannot meet?
The first step is to establish whether the problem lies in the commitment, the resources or the exclusions. Ask for a written recovery plan and a review of the data. If the commitments are unrealistic, it is better to renegotiate them. If the provider fails to honour agreed commitments, consult the contract for the conditions provided and, if necessary, a legal adviser before formally disputing.