Service Level Agreement
A service level agreement for facilities and maintenance, written so it can actually be measured. Four priorities each carrying a worked example, response time and fix time promised separately, and a stop-the-clock section that settles the arguments every other SLA leaves open.
- Four priorities, each with a worked maintenance example
- Response time and fix time promised as two separate things
- Ten stop-the-clock circumstances, each needing evidence
- Service credits banded as a remedy rather than a revenue line
Service Level Agreement
Written to be measured
| # | PRIORITY DEFINITIONS | RESPONSE | FIX | HOURS |
|---|---|---|---|---|
| 1 | P1 Emergency, no heating in winter, water ingress, lift entrapment | |||
| 2 | P2 Urgent, a fault degrading service or likely to escalate | |||
| 3 | P3 Routine, a defect with no immediate service impact |
The document you will get. Download for the full, editable file.
Who this facilities SLA is for
One agreement, and three people who will each read a different section when something goes wrong.
The client or facilities manager
You need the priority examples and the service lines. What you are buying is a response to a lift entrapment at two in the morning, and the examples are what make that unambiguous.
The provider or account manager
The stop-the-clock section is the one that protects you, and the client obligations section is the one that makes the agreement fair. Neither works without evidence recorded against each instance.
Whoever signs the contract
Service credits, their cap and the escalation ladder. Credits are a remedy that triggers a recovery plan rather than a discount somebody budgets for.
Whoever reports performance
Jobs, met, achieved, and the stopped-clock share. That last figure is the one that says whether the achievement number describes the contract anybody signed.
Where the SLA gets tested
The structure suits any maintenance contract. What changes is which priority dominates and how expensive a missed fix is. If you run one of these, the sector page goes further than the template does.
- FM service providers
Contracted delivery, where the SLA is the product and the stopped-clock share is what a client audits first.
FM service provider software - Commercial real estate
Multi tenant buildings, where access to a demised area is the most common legitimate reason a clock stops.
Portfolio maintenance software - Healthcare
Continuity of service, where a P1 is defined by clinical impact rather than by the asset that failed.
Healthcare maintenance software - Retail and malls
Trading hours, which decide both the response window and whether a fix can happen at all before closing.
Retail maintenance software - Corporate facilities
Occupied offices, where comfort calls dominate volume and the priority examples stop them all becoming urgent.
Corporate facilities software - Education
Term time and holidays, where the same fault carries a completely different priority depending on the week.
Campus maintenance software
What a facilities service level agreement should contain
A facilities service level agreement is the written promise of how maintenance work will be responded to and resolved. It defines each priority with a worked example, commits to a response time and a fix time separately, sets a target for every service line with the report that measures it, and states the circumstances that stop the clock.
A. The sections that carry the weight
| Field | What goes in it | Why it earns its place |
|---|---|---|
| Agreement details | The parties, the sites, the term and the review cycle | An SLA with no review date becomes the agreement nobody can change, which is usually the one nobody follows either. |
| Scope, and what is excluded | Services covered, services specifically excluded, and the core hours | Exclusions are the half people skip. An unstated exclusion is read as included by whichever side it suits. |
| How work is requested | The channel, and which one starts the clock | The most common source of response disputes. A job mentioned to a technician on site is not a job raised on the helpdesk, and the agreement has to say so. |
| Priority definitions | Four priorities, each with a worked example and its hours of cover | The example is the part that works. Without one, priority is set by whoever is most annoyed rather than by what is actually happening. |
| Response time | How long until somebody attends or makes contact | A promise about attention. It is the easier promise to keep, which is exactly why it cannot be the only one. |
| Fix time | How long until the fault is actually resolved | The promise the occupant cares about. Separating it from response is what stops a contract being met on paper while the heating is still off. |
| Service levels by service line | Ten lines, each with a target and how it is measured | Every target needs a measurement route beside it. A service level that cannot be produced from a system report will be measured by whoever shouts loudest. |
| Service credits and remedies | Performance bands, the remedy at each, and a cap | Banded so that a first miss triggers a recovery plan rather than a payment. Credits exist to change behavior, not to price failure. |
| When the clock stops | Ten circumstances, each with the evidence required and who agreed it | The section both sides should read twice before signing, and the one no generic SLA template has. |
| Performance this period | Jobs, met, missed and achieved, by priority | Reported per priority rather than blended, because a provider can pass overall while missing every P1. |
| Stopped-clock share | How many jobs had the clock stopped, as a percentage | The integrity check on the whole report. A high achievement figure with a high exclusion rate is not a good month. |
| Escalation and client obligations | Four escalation levels, and what the client has to provide | An SLA with obligations on one side only is not an agreement. Most missed targets have a client-side cause somewhere in the chain. |
The stop-the-clock section is the reason to use this template rather than a generic one, and it is the section a legal SLA template will not have. Maintenance work stops for reasons that have nothing to do with the provider: access to a plant room is refused, a quotation above the agreed limit has not been approved, a part has a stated lead time, a utility or a third party owns the fault, a permit could not be obtained, the work must wait for an agreed shutdown, or the weather makes it unsafe. Without those circumstances written down, every one of them becomes an argument conducted after the fact. Written down, each becomes a recorded instance with evidence and a named person who agreed it. That last requirement matters more than the list. An exclusion that can be applied without evidence is not a stop-the-clock provision, it is a way of making any target achievable, and the stopped-clock share in the reporting section exists precisely so that misuse is visible.
B. What it looks like filled in
One month on a single site. The agreement was met, and the last two lines are the ones worth reading before anybody celebrates.
| Line | Value |
|---|---|
| Jobs in scope | 186 |
| Jobs met | 178 |
| Overall achievement | 96% |
| Contract target | 95% |
| Jobs where the clock was stopped | 14 |
| Stopped clock as a share of jobs | 8% |
Ninety-six against a ninety-five target is a pass, and the stopped-clock line is what tells you whether it is a real one. Fourteen jobs of the one hundred and eighty-six had the clock stopped, which at eight percent is unremarkable and probably honest: on a live site, access gets refused and parts have lead times. The reason the file reports that share at all is that nothing else in the agreement constrains it. A provider consistently achieving target with twenty percent of jobs excluded is reporting on a different contract from the one the client believes they signed, and no amount of scrutiny on the achievement figure will reveal it. Read the two numbers together and always in that order. The other habit worth keeping is reporting by priority rather than blended: this month could show ninety-six percent overall while every P1 was missed, because P1 volume is small and the arithmetic buries it.
Word for the version that goes into a contract or a tender, Excel for the one that calculates achievement by priority and tracks the stopped-clock share with a Performance Log per period, and PDF for issuing to a client or a provider. Free, and yours to rebrand.
How do you write a maintenance SLA?
Write the priorities first, with examples, because everything downstream is measured against them. Then agree what stops the clock before you need it. Six steps.
Set the scope, and say what is excluded
Services covered, services specifically excluded, core hours, out of hours cover and public holidays. Exclusions matter more than inclusions, because an unstated one is read as covered by whoever it suits.
Say which channel starts the clock
Helpdesk, portal, phone or email, and which of them counts as a job being raised. A fault mentioned to a technician in a corridor is where most response disputes actually begin.
Define four priorities, each with a worked example
P1 through P4, with what each means and a real example against it: no heating in winter, water ingress, lift entrapment. Without examples, everything becomes P1 by Friday afternoon.
Promise response and fix separately
Two columns, two commitments, and the hours of cover each applies in. A provider can attend every job inside the response window and resolve none of them inside the fix window.
Set a target per service line with its measurement route
Ten lines from reactive and planned maintenance through to waste and energy, each with a target and the report that produces the number. A target with no report behind it is an opinion.
Agree what stops the clock, and what evidence each needs
Access refused, approval outstanding, parts lead time, third party fault, permit unobtainable, shutdown window, weather. Each with the evidence required and a named person who agreed it.
Response time versus fix time
Two promises that get written as one number, and the gap between them is where an occupant's experience actually lives.
| Aspect | Response time | Fix time |
|---|---|---|
| What it promises | Somebody will attend or make contact | The fault will actually be resolved |
| What it proves when met | The job was picked up | The building is working again |
| Who notices it | The helpdesk and the report | The person who raised it |
| How it is usually missed | Out of hours, or a channel gap | Parts, access or a specialist |
| Why it is the easier one | It needs attention, not resources | It needs the fault to be fixable now |
| What measuring only this hides | Nothing, it is the smaller promise | A contract met with the heating still off |
An SLA that promises only a response is the most common way a maintenance contract looks healthy and feels terrible. Attending within an hour is largely a question of rostering; resolving within four is a question of parts, access and competence, and it is the one the occupant experiences. A provider reporting ninety-eight percent response compliance and nothing about fix is not necessarily hiding anything, but nobody can tell from the report. Promise both, measure both, and report them separately, because the two numbers fail for completely different reasons and only one of them is fixable by trying harder.
When the template starts to feel limiting
The file is built to be agreed once and reported each period. Four things start to hurt as soon as you want it to run itself.
The agreement and the helpdesk are separate
The SLA states the targets and something else records the jobs. Reporting means exporting one and reconciling it against the other by hand, every period.
Stopped clocks are recorded after the fact
A clock stops in the moment and gets written down later, usually when the report is being prepared, which is exactly when the evidence is hardest to reconstruct.
Priority is still set by a person
The examples help enormously and they do not enforce anything. Nothing in a document stops a routine job being logged as urgent because the caller was persuasive.
One file per contract
A provider running several contracts has a workbook each, so a question about P1 performance across clients means opening all of them.
What running this in Facilio looks like
The agreement does not change. What changes is that response, fix and stopped clocks are measured from the work rather than assembled from it.
Work Order Intelligence
Response and fix are timed separately
Both clocks run against the job itself, so the two promises are reported apart without anybody reconciling a helpdesk export against a contract document.
Ops Performance Intelligence
The stopped-clock share is visible
Every exclusion is recorded against its job with its reason, so a rising share surfaces as an exception rather than as a figure nobody thought to total.
Audit Report Intelligence
A challenged month can be opened
Achievement traces back to the jobs behind it, so a disputed figure is answered by showing the work rather than by re-explaining the spreadsheet.
Asset Intelligence
Priority gets a second opinion
Because the asset and its history sit behind the job, a fault on critical plant is recognized as one rather than depending on how the call was described.
Hallucination-free by design. Atom AI answers from the records in your tenant rather than generating plausible text, so an empty field reads as empty rather than filled in for you.
Frequently asked questions
What is a service level agreement in facilities management?
It is the written promise of how maintenance work will be responded to and resolved: what each priority means, how quickly somebody attends, how quickly the fault is fixed, and what happens when a target is missed.
A facilities SLA differs from a generic one in what it has to survive. Access gets refused, parts have lead times and permits cannot always be obtained, so the agreement needs a stop-the-clock section that a generic template will not have.
What is the difference between response time and fix time?
Response time is how long until somebody attends or makes contact. Fix time is how long until the fault is actually resolved. They are two separate promises and should be written and reported as two columns.
Measuring only response is the common failure, because it is the easier promise to keep. A provider can attend every job inside the window and resolve none of them inside the fix target, and a response-only report will look excellent.
How should maintenance priorities be defined?
With a worked example against every level, which is the single most useful thing in this template. P1 covers risk to life, safety, security or total loss of a service, with examples like no heating in winter, water ingress or a lift entrapment.
A priority table with definitions but no examples produces a helpdesk where everything is P1 by Friday afternoon, and the genuine emergencies queue behind requests that were never urgent.
What does stop the clock mean in an SLA?
It is a written list of circumstances where the SLA timer pauses because the delay is outside the provider's control: access denied, spend above the agreed limit not yet approved, parts on a stated lead time, a third party or utility at fault, a permit that could not be obtained, an agreed shutdown window, or unsafe weather.
The critical part is not the list but the conditions on it. Each instance needs evidence and a named person who agreed it, otherwise the exclusion becomes a way of making any target achievable.
How are service credits supposed to work?
As a remedy rather than a revenue line. The template bands them so a first month below target triggers a written recovery plan rather than a payment, with financial credits and a cap beyond that.
An agreement where missing a target is cheaper for the provider than meeting it has been written the wrong way round, and it will produce exactly the behavior it prices.
What is a reasonable stopped-clock share?
Single figures is normal on a live site. The worked example shows fourteen jobs out of one hundred and eighty-six, which is eight percent, and that is the kind of number access refusals and parts lead times genuinely produce.
There is no hard threshold, and that is why the share is reported at all. A provider consistently hitting target with twenty percent of jobs excluded is reporting on a different contract from the one the client thinks they signed.
Does this SLA define the KPIs as well?
No, and keeping them apart is deliberate. This agreement sets targets and the rules around them. How each measure is calculated, including what counts and what does not, belongs in a KPI definition set.
If you need the calculation written out in words for every measure, use the facility management KPI template alongside this. The SLA says what is promised; the KPI set says how the number is worked out.
Should the client have obligations in the SLA too?
Yes, and an agreement with obligations on one side only is not really an agreement. The template asks for access arrangements, information such as asset registers and statutory records, approval turnaround for quoted work, and nominated contacts.
Most missed targets have a client-side cause somewhere in the chain. Naming those obligations is what makes that a conversation rather than an accusation, and it is what gives the stop-the-clock provisions their legitimacy.
In one paragraph
A facilities SLA is only as good as its definitions. Give every priority a worked maintenance example, because a table without them turns every request into an emergency. Promise response and fix separately and report them apart. Put a measurement route beside every target. Then agree what stops the clock, with evidence and a named approver against each instance, and report the stopped-clock share beside the achievement figure so nobody reads one without the other.
The agreement is the floor, not the ceiling
Agree it once, report it every period and watch the exclusion rate as closely as the achievement rate. Once you are reconciling a helpdesk export against a contract document every month, the spreadsheet has done its job.
