SLAs for Cleaning, Security and Maintenance: Metrics to Put in the Contract

Why most SLAs fail before a single metric is set

The service level agreement is the part of a facilities contract that turns 'we will look after your building' into measurable commitments. The ISO 41001 facility-management approach treats service delivery as something to plan, measure, check and improve in a cycle; a contract without measurable service levels cannot support that discipline. The core skill is choosing indicators that reflect the outcome the customer actually wants rather than activity for its own sake.

Vague promises such as 'maintained to a high standard' or 'secure at all times' look reassuring but give neither party an objective basis when a dispute arises. If the contract does not say how cleanliness, patrol quality or repair performance will be measured, by whom and how often, the argument collapses into opinion. Decide on the metric, the data source and the sampling method before you negotiate the target.

• Measure the outcome of the service, not the volume of effort behind it.
• Give every indicator an owner, a data source and a reporting frequency.
• Start with 8–12 core indicators; dozens of metrics mean none is taken seriously.

Cleaning: from 'looks clean' to audited pass rates

Cleaning quality is best defined as a matrix that lists each zone and the tasks to be performed there daily, weekly, monthly or periodically. That matrix is the reference against which any inspection result is judged, so it should be explicit in the contract rather than implied. Building service contractors that operate to a recognised management standard, such as the ISSA Cleaning Industry Management Standard (CIMS), typically build this discipline into documented policies and inspection protocols.

Objective measurement comes from a scored inspection. A practical approach is a per-area checklist where each item passes or fails, producing a percentage for the site or zone. Set a minimum acceptable pass threshold and audit often enough that a failing week is caught immediately rather than discovered a quarter later. A shared dashboard, where client and provider see the same inspection scores and photographic evidence, prevents most disputes because both parties argue from the same numbers.

Cleaning also has a time dimension. Separate routine requests from hygiene incidents such as a slip hazard, an out-of-service restroom or a major spill, and set different response targets for each tier — for example, urgent within one to two hours in working hours, routine within 24. Response speed is one of the most visible signs of service quality for building users.

• Percentage of zones passing the audit at or above the agreed threshold.
• Response and resolution time for urgent hygiene and safety incidents.
• Availability of consumables and restroom condition as separate checklist items.

Security guarding: time, presence and reliability

For guarding, time and presence are the essence of the service. Response to calls and alarms should state the point where the clock starts (such as dispatch or alarm receipt) and the target time to arrive on location. A percentage target — for example, reaching high-priority alarm sites within a stated number of minutes in 95% of events — makes performance measurable rather than anecdotal.

Presence is measured through patrol compliance: completed patrols of defined zones against the schedule for each shift, and through post coverage. Attendance and relief arrangements matter because a post left vacant is the most visible failure a security SLA can produce. Track and report unplanned absences rather than letting them surface only after an incident.

Quality shows up in the accuracy and timeliness of incident reports and in compliance with licensing and training requirements. Because the consequences of poor guarding often appear later, documented competence and supervision checks matter as much as response times. Supervision adequacy and invoice accuracy are also common SLA measures that reveal how the contract is actually run.

• Percentage of priority calls met within the agreed arrival target.
• Patrol completion relative to the schedule for each shift and zone.
• Number of posts left uncovered because of absence.
• Timeliness and accuracy of incident reports and licensing compliance.

Maintenance: reliability, responsiveness and statutory duties

Maintenance metrics split into planned and reactive. The preventive (planned) maintenance completion rate — completed tasks divided by scheduled tasks — shows whether the team is protecting assets or merely reacting to breakdowns. A low score means the crew is spending its time on failures it could have prevented.

For reactive work, track response and completion time by priority, the first-time fix rate, and the classic reliability measures of mean time to repair (MTTR) and mean time between failures (MTBF). For assets whose failure stops the business — HVAC, lifts, power, access control — measure actual uptime or downtime against a stated target and review repeat fault rates to find root causes.

Technical agreements also carry statutory duties: testing of fire systems, lifts, emergency lighting and similar equipment must be completed and documented. Statutory-compliance completion is a hard metric that sits outside ordinary commercial negotiation. Because delayed reporting hides early signs of asset failure and rising cost, most operational indicators should be reviewed monthly, not annually.

• Preventive maintenance completion rate in the scheduled window.
• Response and resolution time by priority, plus first-time fix rate.
• MTTR and MTBF for critical assets, with repeat-fault trend.
• Percentage of statutory inspections and tests completed and documented.

Turning a list of metrics into a workable contract

A metric only works if the contract explains how it is measured. For each indicator define the data source, who records the result, the sampling method (every task, a weekly sample or spot audits), the service window during which the clock runs, and acceptable exclusions. Generic wording such as 'timely service' is weak; 'urgent requests responded to within 30 minutes' is strong enough to act on.

Remedies should be proportionate and evidence-based. The common tool is service credits deducted from the invoice for repeated failures, with an incentive where performance exceeds target. Tie credits to verifiable indicators rather than impressions, set a cap on deductions for a reporting period, and agree a dispute process that refers back to the measurement method.

Agree a cadence — typically monthly — for reviewing scores, trends and corrective actions, and give each indicator an owner and an escalation trigger. The indicators most worth tracking are those that reveal risk before it becomes visible to occupants, such as repeat faults or falling preventive-maintenance compliance. Regular review is what converts an SLA from a document into a management tool.

• Define source, sampler and service window before signing.
• Cap service credits and write a dispute path into the contract.
• Review trends monthly and escalate by exception rather than by complaint.

Common mistakes to avoid when writing the SLA

The biggest mistake is measuring activity instead of outcome: the number of inspections completed matters less than whether inspections reduced faults, complaints and compliance gaps. A second mistake is importing benchmarks from an unrelated site — a hospital, airport, warehouse and office building do not share the same risk profile or operating model. A third is poor data quality: if teams close work orders without accurate notes, every report built on that data becomes unreliable.

Avoid tracking so many indicators that none is taken seriously, and do not defer review to quarterly or 'on complaint' meetings. Delayed reporting hides rising costs, ageing assets and a contractor drifting off standard. Finally, do not make penalties so severe that the provider is incentivised to hide problems rather than fix them; a fair evidence-based regime is more likely to produce honest reporting and genuine improvement.

• Do not copy benchmarks from a different type of building without validation.
• Guard data quality, not just the headline number.
• Agree exclusions in advance rather than after the first missed target.