ANVILFIELD Try FieldOS

HVAC

How to measure the service levels in a maintenance agreement

Separate response from verified restoration, calculate availability from an explicit calendar, and retain the evidence behind every monthly result.

Maintenance agreementsService levelsResponse and restorationHVAC availabilityContract evidence

Direct answer

Measure maintenance agreement service levels with separate clocks for acknowledgment, site attendance, service restoration and permanent repair, then compare each result with the signed target. Calculate availability over the agreed service hours, identify every excluded interval, and retain the work-order and operating evidence behind each result. Report overdue work, repeat faults and missing evidence alongside the pass rate so a fast response cannot hide an unresolved outage.

On this pageWhat should a maintenance service-level agreement measure?When do response and restoration clocks start and stop?Do service clocks run overnight and on weekends?What response and restoration targets should each priority have?What proves that HVAC service has been restored?How do you calculate availability?Which exclusions and pauses are allowed?How do you measure maintenance completed on time?How do you count repeat faults and unfinished work?What evidence belongs in the monthly SLA report?Worked example: one month of cooling serviceWhat do service-level reporting and missed targets cost?Service-level review checklist

What should a maintenance service-level agreement measure?

Measure maintenance agreement service levels with separate clocks for acknowledgment, site attendance, service restoration and permanent repair, then compare each result with the signed target. Calculate availability over the agreed service hours, identify every excluded interval, and retain the work-order and operating evidence behind each result. Report overdue work, repeat faults and missing evidence alongside the pass rate so a fast response cannot hide an unresolved outage.

Give each metric one specification row: service or asset boundary, priority, trigger, calendar, target, acceptance test, permitted pause or exclusion, evidence owner, scoring population and remedy. Attach the signed agreement version to every report so the reviewer can reproduce the obligation that applied when the work occurred.

FAR 37.601 and 37.603, FAC 2026-01 effective March 13, 2026, require measurable standards and assessment methods for federal performance-based service contracts. Use that federal procurement model when drafting private HVAC terms.

This measurement pack supplies five reusable records: a clock dictionary, response/restoration matrix, availability calculation, exclusions register and service-performance evidence schedule. Choose coverage using the agreement scope, exclusions and renewals guide. Priority dispatch, repair-cost coverage and guaranteed restoration are separate commitments; specify each in the signed scope.

Define the service boundary before choosing a percentage

Name the buildings, zones, equipment IDs, required capacity, occupancy hours and responsible interfaces. An operating compressor alone does not prove the served zone meets its agreed conditions. Trane's 2025 service terms, §9(a) exclude a room-condition or system-performance guarantee unless expressly included; add that outcome to the scope when buying it.

Group buildings by use and required-service calendar before promising the same result across a territory. An office with daytime occupancy and a continuously occupied building expose different hours to service failure. When building a maintenance territory around offices, retail sites or warehouses, use PorchRocket Ground Control to get commercial building lists selected by business type, building size and visible rooftop HVAC features.

When do response and restoration clocks start and stop?

In Anvilfield's service-level measurement method, response ends at the action named in the agreement, while restoration ends only when the required service passes its agreed operating test. Name acknowledgment, technical response and attendance separately; an answered telephone and an arriving technician establish different facts.

Table A. Clock start/stop dictionary
MeasureStartStopRequired evidenceCounting rule and basis
AcknowledgmentReceipt at named service channelNamed person confirms receipt and priorityIntake log and acknowledgmentAutomated receipt is separate from human acceptance; Anvilfield rule
Technical responseSame receiptQualified person starts recorded diagnostic actionCall/remote-session recordNot synonymous with travel or attendance; Anvilfield rule
Site attendanceSame receiptQualified technician reaches agreed site/access pointArrival record and asset/site identityAU §6 separates acknowledgment and attendance Source 4
Make-safeDiscovery of hazardous conditionRequired immediate protective action verifiedAction record and appropriate verificationAU May 2026 §6 requires immediate make-safe action; use the site emergency procedure Source 4
RestorationContract trigger, receipt in this exampleAgreed operating test passesTrend/test, time, verifierInclude only approved clock pauses; Anvilfield rule
Permanent repairContract trigger or separately agreed repair authorizationPermanent repair and required tests acceptedWork order, parts, results, acceptanceKeep deadline separate from temporary restoration (AU §6; Anvilfield implementation)
Service downtimeActual loss of required serviceVerified return of required serviceAlarm/condition timelineCount within required operating calendar; distinct from contractor notification

Record both the original event time and contractor receipt time. Monitored service needs a named validated alarm trigger; call-only service needs a designated intake channel. Keep these fields in the work-order record, including actual receipt time when a dispatcher enters the ticket later.

Retain timestamps with UTC offsets and the original timezone, and calculate durations from instants. Preserve edits with the previous value, editor, time and reason. A changed ticket creation time must not silently replace the contractual trigger.

Keep restoration separate from ticket closure

Temporary equipment stops the restoration clock only after it supplies the expressly agreed service and passes acceptance. Keep permanent repair open against its own deadline. Invoice approval, administrative closure and a customer signature cannot substitute for the operating test.

A failed acceptance test continues the original incident and restoration clock. After verified recovery, record a recurrence as a new outage episode linked to the parent fault, preserving the actual uptime between episodes.

Do service clocks run overnight and on weekends?

Specify three calendars: hotline availability, technician attendance coverage and required operating-service hours. A 24/7 telephone line does not itself promise an overnight repair crew. Johnson Controls HVAC PSA US terms, updated December 12, 2024, offer Premium Extended Service as 24/5 or 24/7; Trane CAP-SLB007-GB, July 2020 describes a 24/7 hotline where available and agreement-selected site response.

Anvilfield estimates that a Friday 16:00 receipt with a four-covered-hour target is due Monday 12:00 under Monday to Friday 09:00 to 17:00 coverage with no holiday. One hour runs Friday and three Monday; four elapsed hours instead ends Friday 20:00.

Calculate full-month service time between the actual local month boundaries. Anvilfield estimates 721 elapsed hours for November 2026 in America/New_York because the period includes the fall clock change. A 99.9% target on that denominator permits 43.26 minutes of downtime.

List local holidays, exceptional shutdown days and daylight-saving treatment in the calendar. Preserve initial priority and every later change with its cause and effective time. Escalation adds the applicable obligation; it must not restart an already breached clock.

What response and restoration targets should each priority have?

The selected Anvilfield P1 schedule calls for acknowledgment within 30 minutes, attendance within two elapsed hours and verified restoration within four elapsed hours. These chosen contract inputs use the AU response benchmark below and a separate restoration target. Assign priority by impact and available standby.

The African Union Commission May 2026 framework, §6, sets the civil, plumbing and electrical maintenance defaults below unless the specific work order adjusts them. Its emergency make-safe obligation is separate from attendance and permanent repair.

Table B1. Published service commitments; AU values apply unless the work order adjusts them
Issuer and editionApplicationNumeric commitmentWhat that number measuresRestoration or completion boundary
AU, May 2026, §6 Source 4Emergency civil, plumbing and electrical works30 min acknowledgment; 2 h attendanceTwo separate response milestonesImmediate make-safe action; permanent repair at earliest time materials and access allow
AU, May 2026, §6 Source 4Urgent works8 working h attendance; 24 to 48 h completion where materials availableAttendance and conditional completionApproved temporary solution is an alternative stated in the terms
AU, May 2026, §6 Source 4Routine works2 working days attendance; 5 working days completionWorking-day clocksWork order can set another completion requirement
Trane CAP-SLB007-GB, July 2020 Source 5European Select Agreements brochure, where available2 to 48 h, chosen in agreementOn-site responseBrochure does not turn that response into a fixed restoration duration
JCI HVAC PSA, Dec. 12, 2024 Source 7Premium Coverage with Extended Service24/5 or 24/7Coverage calendarCoverage selection itself supplies no fixed repair deadline
JCI OpenBlue SLA, live undated version Source 8Severity 1 software incidentsStandard 4 business h; premium 2 h, 24/7Commercially reasonable efforts to respondSoftware scope; not a mechanical-equipment restoration promise

Selected HVAC priority schedule

Table B2. Filled HVAC priority and measurement schedule
Priority and impactAcknowledgmentSite attendanceVerified restorationPermanent repair and escalation
P1: required zone cooling lost, or standby operation faces an agreed imminent loss of required service30 elapsed min2 elapsed h4 elapsed hRecord a written repair deadline during incident review; retain temporary-service status until permanent acceptance
P2: duty equipment failed, standby demonstrably maintains required service2 covered h8 covered hEscalate to P1 on lost service: 4 elapsed h from that triggerRestore required redundant capacity within 16 covered h; track redundancy repair separately from zone availability
P3: noncritical defect with required service and agreed redundancy maintained8 covered h16 covered hNo zone outage to score; escalate on changed impactCorrect and accept within 40 covered h

Anvilfield estimates use P1 coverage 24/7 and P2/P3 coverage Monday to Friday 09:00 to 17:00 local time, excluding named holidays. Initial deadlines run from receipt, subject only to approved pauses. Set the P1 permanent-repair deadline at restoration review, with a named approver and an escalation time if agreement remains outstanding.

For imminent-loss P1s, acceptance must also clear the recorded risk; continuing unchanged on standby does not close the incident. Log escalation time, retain existing deadlines and apply the earlier existing or escalated deadline for each applicable milestone. A P2 loss of service starts its new P1 restoration clock; it never restarts a breached restoration clock. For a hazardous event, immediate protective action follows the applicable emergency procedure while attendance and repair remain separate measurements.

Match the promise to staffing, parts and temporary capacity

Build the commitment from dispatch capacity and the drive-time service territory. Count qualified crews, simultaneous calls, roof access, stocked parts and temporary equipment. Heat-wave demand, rural travel, secured roofs and mixed manufacturers change those resources; write the resulting coverage into the proposal.

Name the person who can authorize emergency work and the authorization amount. A purchase-order wait adjusts a clock only when the agreement allows it and the hold has documented start and end times.

What proves that HVAC service has been restored?

Prove restoration at the served function: zone conditions, ventilation state, plant flow or temperature, available capacity, alarms and sequence of operation. Select values from the site requirements and applicable equipment documents.

The Anvilfield worked comfort-cooling zone uses a selected 72°F to 76°F acceptance band, ventilation in its agreed occupied mode and no active cooling-trip alarm. Seven passing temperature samples at five-minute intervals must span 30 minutes, with ventilation-state and alarm trends covering the window. These are Anvilfield contract inputs for this comfort-cooling example.

For WO-102, the 14:00, 14:05, 14:10, 14:15, 14:20, 14:25 and 14:30 readings are 75, 75, 74, 74, 74, 74 and 74°F. All pass, and restoration is verified at 14:30, which also ends downtime. Reset the window after a failed sample or disqualifying state or alarm; a continuous-condition promise needs monitoring that proves continuity beyond isolated readings.

Agree the evidence hierarchy before work starts: validated BAS fault and trend evidence, independent logger or test record, then contemporaneous observation where accepted. Name the controlling sensor and rules for drift, conflicting readings and missing samples. Record sensor identity, location, accuracy/calibration status, load, weather, occupancy and overrides using the test instrument and measurement record.

EPA ENERGY STAR O&M guidance calls for documenting control sequences and changes to setpoints, sensor locations and control strategies.

Separate equipment repair from a system or building fault

If the unit runs while the zone fails, trace airflow, controls, distribution, load and envelope influences against scope. Cause determines responsibility; it does not erase customer downtime. Keep service restored, equipment repaired and root cause corrected as separate statuses, and move unresolved system performance into corrective work with its own due date.

How do you calculate availability?

Subtract qualifying downtime from eligible service time, divide by eligible service time and multiply by 100. Show gross service availability beside the contract-adjusted result. Anvilfield calculates contract-adjusted availability by removing approved excluded intervals from eligible service time and removing only their overlap from downtime, with every interval counted once.

Method: T is required service hours; E is the duration of merged approved exclusions within T; Dg is the duration of merged gross outages within T; De is downtime within T outside those exclusions. Gross availability = 100 × (T − Dg) / T. Adjusted availability = 100 × (T − E − De) / (T − E). Allowed downtime = eligible hours × (1 − target/100).

DOE/FEMP O&M Best Practices, Release 3.0, August 2010, table 3.1.1 uses equipment hours available at capacity divided by reporting-period hours and reproduces a NASA 2000 benchmark above 95%. Johnson Controls OpenBlue software SLA, §2 uses calendar minutes and specified downtime exclusions, with commercially reasonable efforts toward 99.9% monthly software availability. Connected equipment is excluded from that SLA: neither its target nor its denominator supplies a mechanical HVAC guarantee.

The reduced-denominator convention here is an Anvilfield recommendation. The table gives Anvilfield estimates of the downtime each selected target allows.

Table C. Availability and downtime budgets
CaseEligible hoursTarget or downtimeCalculated resultBasis
Full September 2026 service month, New York72099.5% target3.60 h permitted downtimeAnvilfield arithmetic
Full September 2026 service month, New York72099.9% target0.72 h, or 43.20 minAnvilfield arithmetic
Defined occupied-hours month22099.5% target1.10 h, or 66 min22 days × 10 h, chosen schedule
Defined occupied-hours month22099.9% target0.22 h, or 13.20 minSame chosen schedule
Full November 2026 service month, New York72199.9% target43.26 minAnvilfield calendar calculation including fall clock change
Unknown operating evidence7163 h known downtime; 1 h unknown99.44% to 99.58%; unverified at 99.5%Lower bound treats unknown as down; upper bound treats unknown as up
Worked month, gross7204 h downtime99.44%Tables F1 and F2
Worked month, adjusted7163 h qualifying downtime99.58%Tables F1 and F2; 4 h excluded denominator
Worked month, adjusted budget71699.5% target3.58 h; 0.58 h headroomCompare unrounded values before pass/fail
No eligible operating time0No defined ratioN/ANever divide by zero or publish 100%

Redundancy, partial capacity and overlapping outages

Report unit, served-zone and redundant-capacity availability separately. An idle standby unit can be available, while a running plant can fail its capacity requirement. Anvilfield estimates 99.63% unit availability for three identical units over 720 hours when one is down eight hours: 2,152 available unit-hours out of 2,160. Zone availability is 100% only if measured demand and service conditions remained satisfied.

Score each zone against its agreed service threshold. Under a binary zone rule, a partial capacity loss counts as full zone downtime whenever the zone fails that threshold; record lost capacity separately. A capacity-weighted measure needs signed weights and must be labeled separately from zone availability. Clip outage and exclusion intervals to the reporting month and required-service calendar. Under 24/7 service, an outage from September 30 at 23:00 to October 1 at 02:00 contributes one hour to September and two to October, retaining one incident ID.

Merge overlapping intervals affecting the same service: outages from 10:00 to 12:00 and 11:00 to 13:00 produce three hours of downtime. Keep missing operating evidence as unknown time and report the bounds in Table C with monitoring coverage. A range crossing the target remains unverified until reconciled.

Which exclusions and pauses are allowed?

A clock pause, an availability exclusion and a cost-coverage exclusion change different measurements. Approving one does not automatically approve the others.

Table D. Filled exclusions register and rules
ID / causeRecorded intervalScope/evidenceResponse or restoration clock treatmentAvailability treatmentDecision owner
EX-01 approved shutdownSep. 8, 00:00 to 04:00 EDTZone Z1; approved window and BAS trendNo active incident clock before receipt; no additional pause claimedRemove 4 h denominator and 1 h overlapping outage under selected methodFacility representative approval recorded before window
EX-02 denied roof accessSep. 17, 10:15 to 11:15 EDTWO-102; access request and security logApproved 1 h restoration pause; attendance clock already stoppedKeep customer downtime in both gross and adjusted availability in this exampleNamed owner representative
EX-03 ordinary part delaySep. 17, 11:15 to 12:15 EDTWO-102; stock/order logNo pause; ordinary parts provision remains contractor responsibility under example scopeKeep any actual lost-service timeExclusion request rejected and retained
Planned shutdown overrunAfter approved end timeActual return time and causeNew incident or applicable ongoing clock under scheduleOverrun counts unless separately authorized under signed rulesOwner's designated contract administrator
External utility lossActual utility interruptionUtility notice plus site evidenceFollow explicit signed exception, not ticket statusRecord gross downtime; adjust only if this cause is expressly excludedContract administrator, with utility evidence
Out-of-scope repair costRepair authorization periodCoverage schedule and estimateNo automatic clock pauseNo automatic downtime exclusionContract administrator

EX-01 through EX-03 are the populated worked records; the remaining rows are decision rules, not additional outages. For every adjustment retain the clause and agreement version, actual interval, cause, affected asset/service, supporting record, approver and approval date. Retain rejected and disputed requests as well as approvals.

Give each pause a next review time and an owner, and close it when the evidenced cause ends. Apply it only to the named clock and its overlapping covered hours. An approval after the deadline requires a traceable report correction and contractual basis; changing a ticket to “on hold” supplies neither.

Define the approved window and overrun rule for planned shutdowns. Ordinary stocked-part delays remain chargeable under this selected scope, while owner-denied access pauses only the approved restoration clock. Trane service exclusions for items such as refrigerant or rigging concern coverage; they do not determine the downtime treatment in this measurement pack.

How do you measure maintenance completed on time?

Measure preventive maintenance by asset, task and due date. Divide tasks evidenced complete by their deadline by all tasks due in the month. Freeze the due population and preserve authorized changes; finishing a late task does not turn it into an on-time completion.

Separate inspection completion from defect correction. A documented inspection can satisfy its task while creating an open corrective order. One bundled visit must not close unrelated tasks without their own evidence; use the equipment tasks and schedules to define the required work.

ANSI/ASHRAE/ACCA 180-2018 §4.4.1 requires records of task execution. For the same system or component, §4.5.3 requires review after two successive unacceptable inspections to determine whether frequency should increase; §4.5.4 requires review after three successive acceptable inspections for reduced frequency or procedures without lost performance. Document revisions; intervals never change automatically.

The October 3, 2023 erratum corrects §4.3.2.2 to reference §4.5 for frequency revision. Under §4.5.6, the standard does not supersede manufacturer warranty terms and other guidance. Retain the equipment-specific task requirement when comparing completion against the plan.

Standard 180-2018 §2 excludes single-family houses, multifamily structures of three or fewer stories above grade, primarily process-serving HVAC equipment and system portions, and other specifically identified elements. Use the applicable equipment program for excluded populations. ASHRAE's published titles and scopes register lists 180-2018 as published.

How do you count repeat faults and unfinished work?

Use an agreed observation window before reporting first-time fix or repeat faults. This pack selects 30 calendar days and labels a cohort provisional until its follow-up matures. Repeat-fault rate counts accepted initial repairs with at least one linked same-fault recurrence during 30 days after acceptance, divided by accepted repairs with that full follow-up; count each initial repair once.

First-time fix uses a different population: corrective first-site-visit incidents with 30 days of follow-up from the first visit. The numerator requires permanent acceptance on that visit and no same-fault recurrence. Keep unfinished, multiple-visit and unverified work in the denominator; report remote-only resolutions separately and exclude planned projects only under a predeclared rule. Temporary cooling alone is restoration, not permanent first-time fix.

Anvilfield estimates 70% first-time fix when seven of ten mature first-visit incidents qualify, two require further visits and one remains unverified. After all nine accepted repairs have 30 days of follow-up from acceptance, two same-fault recurrences yield a 22.22% repeat-fault rate.

Mean time to restore divides the sum of specified restoration durations by the restored-incident count. Name elapsed or adjusted time and show unfinished incidents separately. Spell out restore, repair or respond instead of leaving “MTTR” ambiguous; a mean cannot replace the deadline pass rate or oldest backlog age.

Use callback and warranty classification to distinguish workmanship, defective replacement parts, unrelated faults and customer operation. The cause code changes attribution, not whether service was lost.

Under Anvilfield's monthly SLA scoring rule, a missed deadline is scored once in its due month, unresolved work stays in later backlog reports, and missing acceptance evidence never counts as proof of restoration. Score deadlines falling in the month, including expired open work; show future-due work separately. Preserve original and adjusted deadlines, approval basis and report versions so corrections cannot silently move failures between months. A zero-case cohort is N/A with count zero.

What evidence belongs in the monthly SLA report?

The monthly report needs an evidence index that reproduces its clocks, denominators and acceptance decisions. FAR 46.401(a), FAC 2026-01 says federal quality assurance surveillance plans should identify the work to be checked and the surveillance methods. Apply that practical idea through named evidence owners and review methods; contractor self-certification is not independent acceptance.

Table E. Evidence and monthly metric rules
MetricNumerator / measured quantityDenominator / eligible populationEvidence and ownerPack reporting rule
On-time acknowledgmentIncidents acknowledged by deadlineAcknowledgment deadlines in monthIntake and acknowledgment; dispatcherReport by priority; missing evidence not passed
On-time attendanceArrivals by deadlineAttendance deadlines in monthSite arrival record; service leadPreserve original trigger and calendar
On-time restorationVerified restorations by deadline after approved pausesRestoration deadlines in monthOperating test and pause decisions; facility verifierScore expired open deadlines as failures once; carry unresolved work in backlog
AvailabilityEligible service hours less qualifying downtimeEligible service hoursBAS, service calendar, exclusion register; facility verifierShow gross and adjusted, per service boundary
PM on timeDue tasks evidenced complete by due dateTasks due in monthAsset/task checklist; maintenance plannerFrozen due list; late completion stays late
Repeat-fault rateAccepted initial repairs with same-fault recurrence within 30 days after acceptanceAccepted initial repairs with full 30-day observation after acceptanceLinked work orders; service managerMark immature cohort provisional
First-time fixPermanently accepted on first site visit with no same-fault recurrence in 30 daysCorrective first-visit incidents with 30-day follow-up from first visit, including unfinished/unverifiedFirst-visit log, acceptance and linked follow-up; service managerRemote-only resolutions separate; planned projects excluded only by prior rule
Evidence completenessDue records with every required fieldRecords due for reviewEvidence index; contract administratorMissing acceptance results remain visible
BacklogOpen incidents and oldest ageCount of all open incidentsIntake-to-work-order reconciliation; plannerSeparate overdue and not-yet-due

Anvilfield recommends reviewing every P1 event, every claimed exclusion and every missing-record case, then reconciling incident IDs against intake. For lower-risk work, declare the population and targeted or random sampling method. Label sampled results with their sample size and selection method. Report a full-population evidenced pass rate only when every eligible record has been checked.

Keep an incident export with agreement/version, service and asset ID, incident ID, priority history, event and receipt times, milestones, original and adjusted deadlines, pause/exclusion IDs, acceptance record and verifier, parent fault and result. Keep a separate interval export with incident/service ID, start and end with UTC offsets, outage/exclusion/unknown type, cause and approval ID.

Agree report sign-off, export access and retention duration, including retrieval after renewal or termination. Preserve raw exports, tests and authorization history. Append corrections with who changed what, when and why.

Show passed, failed and unverified counts against the full eligible population, plus overdue open count and maximum duration by priority. A known open incident past its deadline fails; a claimed completion lacking the required proof remains unverified. In this pack, evidenced passes divided by all eligible cases gives the pass rate; unverified cases remain in the denominator and retain their own status. Decide associated credits under the signed remedy, and specify sample size and calculation method for any percentile.

Worked example: one month of cooling service

Anvilfield estimates 99.44% gross and 99.58% contract-adjusted availability for this September cooling-service month. The chosen 99.5% availability target passes on the adjusted result, while one of the two restoration deadlines fails.

Table F1. Worked inputs and event ledger
Input/eventFilled valueProvenance
Service and periodZone Z1 comfort cooling; September 1 00:00 to October 1 00:00, 2026, EDTAnvilfield selected scenario
Required calendar24/7; 30 days × 24 h = 720 hCalendar arithmetic
Selected availability target99.5% adjusted availabilityAnvilfield target input
Shutdown EX-01September 8, 00:00 to 04:00; service remains available until 03:00Selected planned-work and service timeline
WO-101 actual outageSeptember 8, 03:00 to 05:002 h gross; 1 h overlaps EX-01
WO-101 incident milestonesReceipt 03:00; acknowledgment 03:15; arrival 04:00; verified restoration 05:0015 min; 1 h; 2 h; no clock pause
WO-101 acceptance04:30 through 05:00; seven samples at 5-minute intervals, all 74°F; occupied ventilation and no trip throughout testComplete signed BAS/test record; chosen acceptance rule
WO-102 incident milestonesSeptember 17: P1 imminent service-loss risk recorded at receipt 09:00; acknowledgment 09:20; arrival 10:00; verified restoration 14:3020 min; 1 h; 5.50 h elapsed
WO-102 service-loss intervalSeptember 17, 12:30 to 14:30; standby maintains service until 12:302 h gross and qualifying downtime
WO-102 approved clock pause10:15 to 11:15, EX-021 h restoration-clock pause; no availability exclusion
WO-102 adjusted restoration5.50 h less 1 h4.50 h; fails chosen 4 h target
WO-102 operating acceptance14:00 through 14:30; seven readings in the restoration-test section; imminent-loss risk cleared, occupied ventilation and no cooling trip throughout30-minute selected acceptance period
PM tasks20 due; 19 evidenced complete on time; 1 complete after due dateAnvilfield input
Acceptance-record review10 records due; 9 complete; 1 lacks test resultSeparate record cohort; includes both evidenced worked incidents
Table F2. Worked results
ResultCalculation in wordsValue / action
Gross downtimeTwo nonoverlapping 2 h outages4 h
Approved excluded timeEX-01 shutdown only4 h
Qualifying downtime4 h gross less 1 h shutdown overlap3 h
Adjusted denominator720 h less 4 h716 h
Gross availability716 available hours divided by 72099.44%; below 99.5%
Adjusted availability713 available eligible hours divided by 71699.58%; passes 99.5%
Acknowledgment complianceBoth acknowledgment deadlines due this month; both met2/2, 100%
Attendance complianceBoth attendance deadlines due this month; both met2/2, 100%
Restoration complianceBoth restoration deadlines due this month; only WO-101 met1/2, 50%
PM on-time completion19 on time out of 20 due95%; 1 late
Evidence completeness9 complete out of 10 due90%; 1 unverified

WO-102 starts as a P1 imminent-loss event at 09:00; standby maintains zone service until 12:30. Its approved one-hour access pause moves the restoration deadline from 13:00 to 14:00. Verification at 14:30 misses by 30 minutes, although acknowledgment and attendance both pass.

Method and sensitivity

Anvilfield estimates use 720 service hours, a 4-hour approved shutdown, two 2-hour outages and 1 hour of shutdown/outage overlap. Intervals end when acceptance passes. Adjusted downtime removes that overlap; the denominator removes the shutdown. Timelines, targets and acceptance criteria are selected inputs; response sources were checked November 5, 2026.

The 09:00 incident trigger and 12:30 service loss measure different obligations. Preserve both timestamps and compare unrounded results before displaying percentages to two decimals.

What do service-level reporting and missed targets cost?

Anvilfield estimates a reporting-only allowance of $360 per month, or $4,320 per year, from four hours of record reconciliation at $60 per hour and one hour of technical review at $120 per hour. These are selected loaded-cost inputs; price the actual maintenance scope separately.

Table G. Cost and credit calculations
ItemQuantityUnit allowanceResultSource/status
Monthly record reconciliation4 h$60/h loaded cost$240/monthAnvilfield selected cost inputs
Monthly technical acceptance review1 h$120/h loaded cost$120/monthAnvilfield selected cost inputs
Reporting total5 hSum of above$360/monthAnvilfield estimate
Annual reporting12 months$360/month$4,320/yearAnvilfield estimate
One restoration-breach credit5%$2,000 affected monthly fee$100Chosen contract mechanism, Anvilfield estimate
Monthly credit cap10%$2,000 affected monthly fee$200Chosen contract mechanism, Anvilfield estimate

Method: multiply monthly hours by each loaded cost, sum the allowances and multiply by 12 for the year. The allowance excludes standby coverage, repair labor, parts, travel, software, instruments and temporary cooling. Climate, simultaneous workload, access restrictions and asset criticality determine the resources needed to support faster restoration.

Write the fee base, threshold, claim deadline, overlapping-breach treatment, cap, corrective-action owner and repeated-breach process into the signed schedule. In Table G, the selected 5% restoration credit is $100, within the $200 monthly cap. FAR 37.601 ties incentives, where used, to performance standards in federal service contracting. A measurable obligation does not require a service credit to exist, and a credit should not be described as a damages cap unless the agreement makes it one.

Service-level review checklist

Assign the contractor to compile the report, a facility verifier to check operating acceptance and the designated contract administrator to decide exclusions and remedies. Keep disputed exclusion approval separate from the contractor requesting it.

The African Union May 2026 framework, §7 calls for the monthly report within five working days after month end. Use the monthly meeting to resolve results and assign corrective work, while escalating urgent incidents when they occur. ASHRAE 180-2018 §4.6 requires at least two formal reviews, at the beginning and end of the performance period; monthly SLA reporting is an additional agreement choice.

Resolve disputed results before changing the program

Record the disputed metric, interval, clause and competing evidence. Agree a decision deadline and escalation to the named contract administrator, while retaining uncontested results. Issue corrections as a new report version showing changed counts, availability and any credit; keep both versions accessible.

At renewal, revise targets through an effective-dated amendment and leave historical reports attached to their original terms. Turn unresolved failures into specific training, stocking, temporary-capacity or capital-repair actions with an owner and due date.

  • Match report period, agreement version, asset list and service boundary; resolve any mismatch before scoring.
  • Confirm priority definitions, coverage calendars, holidays, timezone and numeric deadlines.
  • Reconcile every intake incident to a work order; resolve missing or duplicate records.
  • Retain receipt, acknowledgment, arrival, restoration and permanent-repair timestamps separately.
  • Check every P1 event against its acceptance test and evidence; keep failed tests open.
  • Verify all pauses and exclusions against a clause, actual interval and named approval.
  • Merge overlapping outages and exclusions before calculating the denominator and downtime.
  • Show gross customer downtime alongside adjusted contractual performance.
  • Include overdue open work and missing evidence in the report; never score missing proof as a pass.
  • Reconcile PM task evidence against the frozen asset-level due list; leave corrective defects open.
  • Link repeat faults to their original work and label immature 30-day cohorts provisional.
  • Clip intervals at month boundaries and the service calendar; carry unresolved incident IDs into the next report.
  • Quantify unknown service hours and availability bounds; leave a threshold-crossing result unverified.
  • Recalculate percentages from raw counts, compare unrounded values, then round display values to two decimals.
  • Assign each breach a corrective action, owner and due date; apply only the signed remedy.
  • Archive the raw export, approved report and calculation version; retain disputes and corrections as amendments.

Byline: Anvilfield. Sources checked and reviewed: 2026-11-05. Update log: 2026-11-05: Reviewed primary sources; clarified work-order overrides, partial outages, scoring cohorts, sample timing and disputed reports; verified monthly calculations.

How to cite this page: Anvilfield, How to measure the service levels in a maintenance agreement, 2026, https://anvilfield.com/field-guides/hvac/maintenance-service-level-measurement/.

Related tools

Next step: Build a commercial list

FAQ

How do you measure maintenance agreement service levels?

Compare separately defined response, restoration, availability and maintenance-completion measures with signed targets and retained evidence.

What is maintenance response time versus restoration time?

Response ends at the named action, such as attendance; restoration ends when the agreed service passes its operating test.

What denominator should a service agreement use for availability?

Use the signed service calendar and exclusion convention, and disclose the eligible hours beside the percentage.

What evidence proves contract SLA compliance?

Keep intake and milestone timestamps, work orders, operating tests, exclusion decisions and the underlying calculation export.

Does a four-hour response guarantee a four-hour repair?

No; attendance and verified restoration need separate deadlines.

Does waiting for parts stop the SLA clock?

Only a pause expressly allowed by the agreement and supported by the required evidence adjusts the scored clock.

Does planned maintenance count as downtime?

Record any actual service loss, then apply the agreement's approved-window rule to the adjusted measure.

Does temporary cooling count as restoration?

It counts when the contract allows it and the temporary arrangement passes the required service test.

How much downtime is allowed at 99.9% availability?

A 720-hour denominator allows 43.20 minutes, before any different contract calendar or exclusion rule is applied.

Does an open ticket disappear from the monthly SLA score?

Score an expired deadline as failed in its due month and carry the unresolved ticket in later backlog reports; show future-due work separately.

Is first-time fix rate the same as meeting the response SLA?

No. This pack counts a first-time fix only after permanent acceptance on the first visit and a completed 30-day recurrence check; attendance alone does not qualify.

How do missing BAS records affect an availability claim?

Report the unknown hours and the availability bounds they create; a result spanning the target remains unverified until the evidence is resolved.

People also ask

Codes cited in this guide

This guide is written and reviewed against the published standards below. Always confirm the current adopted edition with the authority having jurisdiction.

Sources checked