Service Level Agreement (SLA): What to Include and the Traps to Avoid
A Service Level Agreement (SLA) is a documented, agreed standard of service between two parties — most often between a contact centre (or its outsourcer) and the business it serves. It sets out what will be delivered, how it is measured, and what happens when the agreed standard is missed.
An SLA is what turns a vague expectation like “answer the phones quickly” into a specific, measurable commitment. It becomes the reference point for staffing, reporting, and holding a provider — or another department — to account.
But an SLA is only ever as good as the behaviour it drives, and that is where most of them quietly go wrong. This guide covers what a good SLA contains, the difference between internal and external SLAs, and why hitting the number is not the same as helping the customer.
Why it matters
An SLA converts expectations into an accountable commitment — the basis for staffing, supplier management and reporting in any contact centre.
Where it gets tricky
Attach targets to speed alone and people optimise speed, not outcomes. The SLA goes green while the experience gets worse.
What this guide covers
The definition, SLA versus service level versus KPI, internal versus external, what to include, the target myth, and how to set SLAs that hold up.
What is a Service Level Agreement?
A service level agreement is a formal, documented commitment that defines the level of service one party will provide to another. It states the standards to be met, how performance will be measured, who is responsible for what, and the consequences if the agreed standard is not delivered.
SLAs exist wherever a service is provided against an expectation — IT support, logistics, facilities — but in the contact centre world they most commonly cover how quickly and how well customer contacts are handled across phone, email, chat and other channels.
In plain English
An SLA is a written promise about the standard of service: here is what we will deliver, here is how we will measure it, and here is what happens if we do not.
It is the difference between “we will answer quickly” and “we will answer 80% of contacts within 20 seconds, measured monthly, with a service credit if we fall below 75%.”
✓ An SLA is
- A measurable standard of service, agreed between two parties
- Specific about metrics, targets, measurement and remedies
- A tool for accountability — internal or external
- A living document, reviewed as the operation changes
✗ An SLA is not
- The same thing as Service Level (that is one metric inside it)
- A guarantee of a good experience — only of a measured proxy
- A set-and-forget contract clause
- A stick to beat a provider with and nothing more
SLA vs Service Level vs KPI
These three terms get used interchangeably in everyday conversation, but they describe different things — and mixing them up is how SLAs end up vague or unenforceable.
The SLA
The whole agreement — the document that bundles the metrics, targets, measurement rules, responsibilities and remedies into one accountable commitment.
Service Level
A single metric: the percentage of contacts answered within a target time, usually written as 80/20. For voice specifically this is also called Grade of Service (GOS). It is one line inside the SLA, not the SLA itself.
The KPIs
Any Key Performance Indicator you could track — Average Speed of Answer (ASA), Abandonment Rate, First Contact Resolution, Occupancy. An SLA is built from a chosen few, with targets attached.
The simplest way to hold it
Your Call Centre KPIs are everything you could measure. The SLA is the short list you have agreed to be held to, with targets and consequences. Service Level is just the best-known KPI that usually sits inside it.
Internal vs external SLAs (you need both)
Most people think of an SLA as something between a business and an outside provider. That is the external SLA, and it is only half the picture. The more overlooked half is the set of internal SLAs between departments — and that is usually where customer promises quietly fall apart.
External SLA
- Between a business and an outside party, typically a client and a BPO (Business Process Outsourcing) provider
- Contractually binding, with commercial consequences
- Commonly includes service credits, earn-back mechanisms and exit triggers
- Needs airtight measurement definitions — the money makes the detail matter
Internal SLA
- Between the contact centre and the departments it depends on — product, IT, billing, fulfilment, legal
- Usually a management and expectation-setting tool rather than a contract
- Remedies are operational — escalation, remediation, re-prioritisation
- Sometimes formalised as an OLA (Operational Level Agreement) — the internal commitments that make the external promise achievable
The chain is only as strong as its weakest internal link
Your external promise to the customer sits on top of a chain of internal handoffs — and the customer only ever experiences the slowest link, not your headline target.
Say your Customer Complaints Charter promises to resolve every enquiry within seven days. The front line does its part in hours. But some of those cases need the product team to answer an internal email, and they routinely take two weeks, because nobody ever agreed an internal SLA with them. The seven-day promise is broken before the agent even picks up the case, and no amount of front-line effort can fix it.
This is why mature operations do not just sign external SLAs — they back them with internal ones. Every customer-facing commitment is traced back through the departments it relies on, and each of those gets its own agreed response standard well inside the customer-facing window.
The practical test
For every customer-facing target, ask: which internal teams does this depend on, and have they agreed a turnaround that fits inside it? If the answer is no, you do not have an SLA — you have a hope.
Why SLAs matter
A good SLA does different jobs for different people. Done well, it aligns them; done badly, it sets them against each other.
For CX leaders
The SLA is where the customer promise becomes measurable. It is the mechanism that stops “we care about customers” from being a poster on the wall and makes it something you can actually report against.
For contact centre leaders
The SLA drives the operating model — the staffing, the rostering, the escalation paths. Get the target right and it is a planning tool; get it wrong and it is a stick you will be beaten with.
For operations and finance
With an outsourced provider, the SLA is the commercial control: service credits, earn-backs and exit rights. It is also the number that determines how much resource the service actually requires.
What goes into an SLA
A robust SLA is more than a target number. These parts separate an agreement you can manage against from one that falls apart the first time performance is disputed.
Scope and services
Exactly which contacts, channels and hours are covered — for example inbound voice and live chat, business hours only, excluding public holidays. Ambiguity here is where most disputes begin.
Metrics and targets
The specific measures and the number attached to each — Service Level, ASA, quality score, resolution. Fewer, meaningful targets beat a long wish-list nobody manages to.
Measurement method
How each metric is calculated and from what source. Does the Service Level clock start at connection or after the IVR? Are very short abandons excluded? Define it, or you will argue it.
Reporting and cadence
What is reported, how often, and over what window — interval, daily or monthly. A monthly average and a daily one tell very different stories about the same performance.
Roles and responsibilities
Who does what — including the client’s obligations, such as accurate forecasts, timely approvals and system access. SLAs fail on both sides, not just the provider’s.
Remedies, exclusions and review
What happens when targets are missed, what is fairly excluded (outages outside the provider’s control, client-caused spikes), and how the SLA itself is reviewed as volumes and channels change.
How to set SLAs well
Good SLAs are short, honest and resourced. This order keeps them useful rather than decorative.
Start from the customer outcome
Decide what a good experience looks like first — resolved, low-effort, in reasonable time — then choose metrics that reflect it. Metrics chosen before outcomes end up measuring convenience, not value.
Pick few, meaningful metrics
Three or four targets that matter beat a dozen nobody manages to. Every extra metric dilutes focus and adds another way to look busy while missing the point.
Define the measurement precisely
Write down exactly how each metric is calculated, from which system, over which window, with which exclusions. Undefined measurement is the single biggest source of SLA disputes.
Set realistic, resourced targets
A target you have not staffed for is a target you have chosen to miss. Model the Service Level you are promising against the headcount you actually have before you commit to a number.
Back it with internal SLAs
Trace every customer-facing target to the internal teams it depends on and agree a turnaround with each that fits inside the promise. An external SLA without internal ones is built on sand.
Agree reporting and a review cadence
Fix how and how often performance is reported, and schedule regular reviews. Volumes, channels and expectations move, and a static SLA slowly stops describing reality.
The benefits of a good SLA
When it is built around outcomes rather than box-ticking, an SLA earns its place. Here is what a well-designed one actually delivers.
Clear expectations
Everyone knows what “good” means, in numbers, so debates become about performance rather than about what was promised.
Accountability
Performance is measured against an agreed standard, with defined consequences — internally and with any external provider.
A basis for staffing
A resourced Service Level target is the input to workforce planning — the link between the promise and the roster.
Aligned departments
Internal SLAs join the front line to the teams behind it, so the customer-facing promise is actually deliverable end to end.
Better decisions
Consistent measurement over time surfaces trends — where performance is slipping and where more resource genuinely pays off.
Healthier supplier relationships
A fair, well-defined SLA with balanced remedies builds a working partnership, not a monthly argument about who caused what.
Common metrics and the target myth
Ask what a “good” SLA target is and someone will quote you 80/20 — 80% of calls answered in 20 seconds — as if it were a law of nature. It is not. It is a convention that dates back to a single early study, and treating it as a universal standard is one of the most common mistakes in the industry.
There is no universal correct target
The right Service Level depends on your customers, the value of each contact, your channels and your cost model. A premium service and a high-volume transactional line should not carry the same target, and copying 80/20 because everyone else quotes it is not a strategy.
The same caution applies to the other metrics people reach for. Abandonment Rate, in particular, is a secondary outcome — it flows from your Service Level performance and customer behaviour, it is not something agents control, and a number like “under 5%” is meaningless quoted on its own without the Service Level context that produced it.
✓ Metrics worth agreeing
- Service Level or ASA — accessibility
- First Contact Resolution — whether the issue was actually solved
- Quality or compliance score — how well, not just how fast
- CSAT or customer effort — the customer’s own verdict
✗ Metrics to handle with care
- Average Handling Time as a hard target — drives rushed, unresolved contacts
- Abandonment Rate as a KPI target — it is an outcome, not a lever agents pull
- Averages with no tail measure — they hide the worst experiences
- Too many metrics at once — focus dilutes accountability
Common SLA mistakes
Most SLA disappointment traces back to a handful of avoidable traps.
Chasing a generic benchmark
Adopting 80/20 (or any other number) because it is “industry standard” rather than because it fits your customers and cost model. A borrowed target is a guess wearing a suit.
Speed with no resolution measure
An accessibility target on its own rewards answering fast, not solving the problem. Without a resolution or quality measure beside it, the SLA can be met while the customer leaves unhappy.
Averages that hide the tail
Reporting only monthly averages masks the customers who had the worst experience. The average can be green while a meaningful minority waited far too long.
No internal SLAs
A customer-facing promise with no agreed turnarounds from the departments it depends on. The front line carries a target it cannot actually control.
Unrealistic, unresourced targets
Promising a Service Level the staffing cannot support just guarantees failure and finger-pointing. Targets have to be modelled against real capacity.
No agreed measurement definition
If both sides calculate the metric differently, the SLA is unenforceable. Define the clock, the source and the exclusions before anyone signs.
How to know your SLA is working
A green dashboard is not proof the SLA is doing its job. These checks tell you whether the number actually reflects a good experience.
Watch the tail, not just the average
Track maximum wait and the distribution alongside the headline. An 80/20 monthly average can hide the one-in-ten customer who waited fifteen minutes.
Read speed and resolution together
If Service Level is up but First Contact Resolution is down, the SLA is being met at the customer’s expense. The two numbers only mean something side by side.
Check the internal links are holding
Monitor the internal SLAs behind the promise. A rising number of cases stuck waiting on another department is an early warning the customer-facing target is about to slip.
Review and adjust on a cadence
Revisit the SLA on a fixed schedule against current volumes and channels. A target that made sense a year ago may now be measuring the wrong thing.
The rule of thumb
Never run an accessibility SLA — Service Level, ASA, Abandonment Rate — without a resolution or quality measure beside it. Speed without resolution is just a faster way to disappoint people.
Frequently Asked Questions
Is an SLA the same as Service Level?
No. Service Level is a single metric — the percentage of contacts answered within a target time, such as 80/20. An SLA is the whole agreement, which may include Service Level along with other metrics, targets, measurement rules and remedies.
What is a good Service Level target for a call centre?
There is no universal right answer, and anyone who quotes you 80/20 as a standard is repeating a convention, not a rule. The correct target depends on your customers’ expectations, the value of each contact, your channels and what you can afford to staff. Set it deliberately, and model it against your capacity — do not copy it.
What is the difference between an internal and an external SLA?
An external SLA is the promise to an outside party, such as a client or a BPO provider, and is usually contractual. An internal SLA is between departments inside the business — the contact centre and product, IT or billing, for example. Internal SLAs are what make an external promise deliverable; the customer experiences the slowest internal link, not your headline target.
Should a contact centre have internal SLAs with other departments?
Yes, and it is one of the most neglected parts of service management. Any customer-facing target that depends on another team is only deliverable if that team has agreed a turnaround that fits inside it. Without internal SLAs, a seven-day resolution promise can be quietly broken by a department that takes two weeks to reply to an internal request.
Are SLAs legally binding?
It depends on the SLA. An outsourced SLA that forms part of a commercial contract is generally binding, with remedies such as service credits. An internal SLA between teams is usually a management tool rather than an enforceable contract. This is general information, not legal advice — have commercial agreements reviewed by a qualified professional.
What is the difference between an SLA and an OLA?
An SLA is the external-facing promise. An OLA — Operational Level Agreement — is an internal agreement between the supporting teams that make the SLA achievable. If IT owns the phone system, an OLA might commit them to a fix time that lets the contact centre meet its SLA.
What penalties are typical in an outsourced SLA?
The most common mechanism is a service credit — a reduction in fees when targets are missed — sometimes with earn-back provisions if performance recovers. Repeated or severe breaches can trigger remediation plans or, ultimately, exit rights. The specifics are commercially negotiated.
Should SLAs cover digital channels like chat and email?
Yes, but with channel-appropriate metrics. Email suits a resolution-time target, live chat suits a speed-to-answer target similar to voice, and asynchronous messaging needs its own definition. Applying a voice SLA unchanged to every channel usually measures the wrong thing.
Can an SLA actually hurt customer experience?
Yes. If you attach targets and penalties to speed metrics alone, people optimise for speed — short-calling, early transfers, pushing unwanted self-service — while resolution and satisfaction fall. Always pair accessibility metrics with resolution and quality measures so the SLA reflects the outcome, not just the proxy.
Where to next
Looking for an outsourced provider?
If your SLA is going to sit with a partner, browse providers in the ACXPA Supplier Directory and filter by location. Front-office covers customer-facing work — contact centres, customer service and sales; back-office covers behind-the-scenes processing such as data, admin and finance.
Summary: Service Level Agreements
A service level agreement is a documented, measurable promise about the standard of service — what is delivered, how it is measured, and what happens when it is missed.
It is not the same as Service Level (a single metric) or a KPI (any measure). The SLA is the short list of metrics you have agreed to be held to, with targets and remedies attached, defined precisely enough that both sides calculate them the same way.
SLAs come in two forms that both matter. The external ones, with clients and providers, are only ever as deliverable as the internal ones between departments — because the customer experiences the slowest internal link, not your headline target.
And the trap worth remembering: an SLA measures a proxy for good service. There is no universal correct target, 80/20 is a convention rather than a standard, and speed attached to penalties on its own just teaches people to game it. Pair every accessibility metric with a resolution or quality measure, watch the tail as well as the average, back the promise with internal SLAs, and staff for the targets you set. Do that and the SLA does its real job — turning a vague expectation into an accountable, honest commitment that protects the experience it was meant to describe.
