Experience Level Agreements: A Complete Guide

An experience level agreement measures what using a service feels like, alongside the service level agreement that measures whether the provider met its commitments. The two diverge constantly, and the gap is not a measurement error — it is the finding.

ITIL 5 treats experience as its own discipline rather than something distributed across practices. This guide covers what an experience level agreement should contain, which metrics actually work, and how to avoid the failure modes that turn XLAs into another dashboard nobody reads.

Why an experience level agreement exists at all

Every service management function has met the situation. Availability is 99.95%. Every response target is green. The service desk hits its resolution times. And the business thinks IT is slow.

Nobody is lying. The SLA measures conformance to what was agreed. It does not establish that what was agreed was the right thing, and it does not measure the parts of the experience nobody wrote down: how much effort the user spent, how long they waited between steps that each met their target, whether they understood what was happening.

An XLA is the instrument for those. It commits the provider to something about the experience, measured from the consumer’s position rather than from internal instrumentation.

What goes in an experience level agreement

An XLA is a short document. Padding it defeats the purpose.

• The service and the consumer group it covers. Experience differs by audience; one agreement covering everyone measures nobody.
• What the consumer should be able to expect, stated in their words. If this section can be satisfied by a green dashboard, it is written wrongly.
• The experience commitments, each with a metric, a target and the performance metric it is read against.
• How each metric is collected, at what frequency, and who owns it.
• What happens when a commitment is missed.
• The review cadence and the parties.

The pairing in the third item is the design decision that makes the whole thing work. An experience metric with no paired performance metric cannot be diagnosed — you know satisfaction fell, and nothing about why.

Metrics that work in an experience level agreement

Eight metrics cover most services. Each is paired with the performance metric it should be read against.

Experience metric	What it measures	Read against
Time to productive use	Request to the user actually working, as they experience it	Request fulfilment within target
Effort score	How hard the user had to work to get the outcome	First-contact resolution rate
Perceived availability	Whether the service was there when needed	Measured availability
Perceived speed	Whether the service felt fast enough for the task	Response time percentiles
Communication quality	Whether users felt informed during disruption	Major incident communication cadence
Confidence	Whether users trust it to work next time	Change failure rate
Findability	Whether users can find the service and its support route	Catalogue completeness
Sentiment trend	Direction of unsolicited feedback over time	Satisfaction score

Two are worth singling out. Effort score is more actionable and more stable than satisfaction — people are poor at rating how satisfied they are and good at reporting how hard something was. And time to productive use is the single most revealing journey metric, because it spans every step and every wait between them, which is exactly what per-ticket metrics hide.

Principles that keep XLAs honest

Measure the journey, not only the transaction. A new-starter provisioning journey where every ticket met its SLA and the person could not work for three days is a real outcome that transaction metrics cannot express.

Ask few questions, often. A single question after a real interaction beats an annual survey nobody completes. Response rate is itself an experience signal — a falling one is evidence, not noise.

Watch unsolicited feedback. Complaints, channel chatter and abandonment carry signal that surveys miss, because the most dissatisfied users stop responding to surveys first.

Close the loop. Tell users what changed because of feedback they gave. Feedback collected and never acknowledged reduces future response, which makes the next measurement worse. This is the most commonly skipped step and the most damaging.

What happens when a commitment is missed

A missed experience commitment should not normally be a service credit event unless explicitly agreed. Financialising experience produces gaming rather than improvement.

A proportionate response ladder:

• A single-period miss — recorded, cause investigated, within one reporting period.
• Two consecutive misses — an improvement item raised in the improvement register.
• Persistent divergence from the paired performance metric — a joint review with the consumer representative. This is the interesting case: performance is fine and experience is not, which usually means the SLA measures the wrong thing.
• A commitment found to be unmeasurable — the agreement is amended, not quietly dropped.

Building the first one

The common mistake is starting with the agreement. Start with the measurement instead, run it for a quarter unattached to any commitment, and only then write down what you are willing to promise.

A workable sequence:

• Pick one service with identifiable users. Not the most troubled one — one where the relationship is good enough to survive an honest number.
• Choose two metrics. Effort score and time to productive use cover most ground for a support-heavy service.
• Instrument them and say nothing. A quarter of baseline before any commitment is what stops you promising a target you cannot hit.
• Show the paired performance metric next to each. This is the conversation that changes minds — a room looking at 99.95% availability beside 61% perceived availability stops arguing about whether the number is fair.
• Write the commitment last, at a level the baseline says is achievable, with the response ladder agreed up front.

Expect the first quarter to be uncomfortable. Services that have reported green for years rarely score well the first time somebody asks the users, and a provider that treats that as an attack rather than as information will not get a second honest data set.

The failure modes

Three ways an XLA programme goes wrong, all avoidable.

Adjusting the survey instead of the service. When experience scores disappoint, the temptation is to reword the question or change the sampling. If the gap between experience and performance is persistent, that gap is the finding — raise it as an improvement item rather than tuning it away.

Measuring what is easy. Provider instrumentation is abundant and cheap; consumer perception is neither. An XLA built entirely from telemetry has quietly become a second SLA.

Using experience data to manage staff. The fastest way to destroy the honesty of the data. Experience measurement is about the service, and data about identifiable individuals carries data protection obligations besides.

Where XLAs fit in ITIL 5

Experience runs through the service journey — the whole of a consumer’s interaction from first awareness of a need through to exit — rather than sitting inside one practice. The onboarding stage in particular is where time to productive use is won or lost, and it is rarely instrumented.

Service level management remains the practice that owns the agreements. An XLA does not replace the SLA; it sits alongside it, reported together so the divergence is visible rather than reconciled away. Our overview of what changed in ITIL 5 covers where experience sits in the edition, and the deliver activity is where most of this is measured. PeopleCert’s Foundation material covers the service journey model directly.

Frequently asked questions

Does an experience level agreement replace the SLA?

No. It sits alongside it. The SLA commits to performance; the XLA commits to experience. Report them together — the gap between them is the most useful thing either produces.

How many commitments should an XLA contain?

Three or four. An XLA with a dozen commitments is a dashboard, and it will be reported rather than acted on.

Can we run XLAs without a survey tool?

Yes. A single question prompted after a real interaction, plus existing records for timing metrics, covers most of the eight metrics above.

What if experience scores stay poor while performance is green?

That is the finding, not a fault in the measurement. It usually means the service levels were agreed against provider-visible measures rather than consumer outcomes.

Who owns the experience level agreement?

The service owner, with a named consumer representative on the other side. Without a counterpart, an XLA is a self-assessment.

Templates for experience management

Our ITIL Toolkit includes an experience management policy, an experience level agreement template with the commitment and response structures described here, and an experience metrics catalog pairing each experience metric with the performance metric it should be read against — within 57 templates aligned to the current edition.