Buying Confidence, Not Clock-Watching: Redrafting MSP Service Levels Around Results

A managed service provider can meet every service level in its contract during a quarter in which the business it supports has a genuinely poor three months. That is not a scandal. It is what happens when the agreement measures reflexes rather than results.

The comfortable service level

Most managed IT contracts signed by UK SMEs still describe the supplier's responsiveness: how quickly the phone was answered, how fast a first response landed, what proportion of tickets met their target, how much of the month the servers were up. Each is a legitimate operational measure. None of them is an outcome.

Consider the modest example of a password reset. Answered in four minutes, closed in twelve — a clean pass against any reasonable target. For the person locked out, it is still a twenty-minute interruption in the middle of a client call, and if it happens twice a week to forty people, the aggregate loss is real and entirely invisible on the service report.

That invisibility is the core problem. An activity-based service level tells you how hard the supplier worked. It does not tell you whether the business ran.

Why the familiar metrics flatter everyone

A few structural quirks are worth naming plainly.

• Closure rewards closing. A ticket marked resolved is a success, even where the underlying fault recurs a fortnight later. Reopened tickets are frequently excluded from attainment calculations.
• First response is easily automated. An acknowledgement from a ticketing system counts as a response in more contracts than buyers realise.
• Uptime is measured where the supplier can see it. A server can be available while the application sitting on it is unusable for half the staff.
• Resolution is defined by the party being measured. Without a user confirmation step, the supplier decides when the clock stops.
• Improvement work has no service level at all. The project to fix the recurring problem is typically a separate statement of work with its own commercial pressures and no public score.

The result is a contract that reports favourably and informs poorly.

Four outcomes worth buying

An outcome-based schedule need not be elaborate. Four categories cover most of what a UK SME actually wants from a managed service.

User productivity. The aggregate time staff could not work because of technology, expressed in hours and split by team or site. Imperfect, yes — but a figure of 180 lost hours in a month is a management fact, where "96 per cent of tickets within target" is not.

Recovery confidence. Not backup success rates, which measure jobs rather than recoverability, but documented end-to-end restores with elapsed time compared against an agreed recovery objective. One evidenced restore per system per year is a reasonable floor for most SMEs.

Change delivery. The scheduled improvements — the new starter process, the patch, the migration, the fix for the thing that keeps breaking — delivered on the date promised. This is usually the weakest area of any contract and the one with the clearest business value.

Reporting quality. Whether the figures can be audited and exported. If a buyer cannot pull the raw ticket data and check the summary, the summary is marketing.

Clauses that move the conversation

Sample wording, to be adapted by a solicitor, might run as follows.

"The Supplier shall report monthly the aggregate hours of user productivity lost due to incidents and service requests in scope, calculated from ticket open and close timestamps and validated against the outage log."

"The Supplier shall demonstrate a successful end-to-end restore of each in-scope system at least once per contract year, evidencing elapsed recovery time against the agreed recovery time objective, and shall provide the test record to the Customer on request."

"The parties shall agree a schedule of improvement works at the start of each quarter. Delivery against that schedule shall be reported as a primary service level, with any date change requiring written agreement."

"A ticket may be recorded as resolved only with user confirmation, or after five working days without recurrence. Tickets reopened within ten working days shall count against the original service level."

"Service credits shall be calculated by reference to the outcome measures in Schedule 3, and shall not be limited to response and resolution times."

Two further protections are worth adding: a right to export the full ticket audit trail, and a defined review meeting at which the outcome data is discussed jointly rather than simply published.

Questions to ask before signing

• Who decides when a ticket is resolved, and can we change that definition?
• How would you measure the hours our staff lost last month, and can we see the working?
• When did you last restore a system end to end for a client, and how long did it take?
• What happened to the improvement list we agreed twelve months ago?
• If an outcome target is missed, what changes in your behaviour — beyond a credit note?
• Can we have a read-only export of the ticket data, today?

A supplier who can answer these without discomfort is demonstrating the thing the buyer is actually trying to purchase.

The honest trade-offs

Outcome measures are harder to draft, cost more to instrument and occasionally provoke argument about attribution — an outage caused by a third-party application is not straightforwardly the provider's fault. Nor should activity metrics be discarded. Response times still matter; they simply belong one layer beneath the outcomes, as diagnostics rather than headlines.

Buyers should also expect the commercial model to shift. A supplier asked to carry productivity and change-delivery risk will price for it, and may need to invest in telemetry it does not currently run. That investment is often the point: it is the difference between a provider that watches a queue and one that understands a business.

What you are actually buying

A service level agreement is a description of what has been purchased. If it describes only how quickly someone picks up the phone, that is precisely what the buyer will receive — promptly answered calls, and no clear answer to whether the business had a good quarter.

For UK SMEs renewing in 2026, the practical step is modest. Take the existing schedule, add one outcome measure — productive time lost is the simplest to instrument — and agree the calculation method with the supplier before the next renewal date. One measure, honestly calculated, will teach a board more than a decade of green dashboards.