Facility management KPIs and SLAs: what to measure and what to ignore

Most facility service level agreements fail because they measure what is easy to log rather than what the buyer needs to know. Three things cause almost all of it: input measures dressed up as outcomes, more KPIs than anyone can act on, and no consequence attached to a miss. A green scorecard, a forty-minute review, and a washroom still out of consumables on the Thursday you walk the floor.

This is the working detail behind the service-levels summary in the buyer’s guide to facility management services in India. The category itself is covered in what integrated facility management is, and the money in how an FM contract is priced.

Key takeaways

• Every measure is an input, an output or an outcome. Only the last two belong in a scorecard that carries money.
• Attendance is an integrity check, not a service measure: it proves you pay for the roster, not that the work was done.
• Response, restoration and resolution are three clocks. Severity turns on consequence, not on how the fault is fixed.
• A credit larger than the provider’s monthly margin stops being a penalty and becomes a dispute.
• A score filled in by the person who did the work is not evidence.

Three reasons the schedule fails

A scorecard fails in one of three ways: an input dressed as an outcome, more measures than anyone can act on, or no consequence attached to a miss.

Failure	What it looks like	What replaces it
An input dressed as an outcome	“Attendance 98%”, “staff deployed as per roster”	Completion against the dated schedule, inspected
Too many measures	Forty KPIs, none discussed	One output and one outcome per service family
No consequence	A miss minuted and carried forward	A severity clock, a credit, escalation on recurrence

Input, output and outcome measures

An input measures what you put in, an output measures what was delivered against a standard, and an outcome measures what users experienced. Headcount and shifts rostered are inputs; areas cleaned against the schedule and AMC visits are outputs; complaint rate by area, temperature excursions and DG take-load are outcomes.

Attendance is an integrity check, not a service measure

Attendance answers a narrow question: are the people you pay for present, on the shift you bought? A contract overstaffed on paper and understaffed at 8am is one of the most expensive failures in this sector, so check it monthly. It is still not evidence about the service.

Score it as a service KPI and the provider optimises that number alone: the roster gets reconciled to the clock rather than the building, and credit rides on a figure that says nothing about the chiller. Treat it as a gate instead. If deployment falls short, the shift is made good and the invoice adjusted for the person-days not supplied, with the wage evidence checked on the same visit, as the contract labour compliance guide sets out.

A KPI catalogue by service family

The catalogue is raw material, not a scorecard.

KPI	What it actually measures	Data source	Cadence
Cleaning: tasks done against the schedule	Whether scheduled work happened where scheduled	Planner against completion log	Daily, monthly
Pantry and cafeteria: window uptime and stock-outs	Whether the counter was open and stocked	Pantry log, stock count	Daily
Pest control: treatments done, sightings after	Whether the cycle ran and activity recurred	Vendor reports, reception register	Per cycle
HVAC: availability and excursions	Whether the plant delivered the condition bought	BMS trends, log book, breakdown register	Daily, monthly
Electrical and DG: take-load and ride-through tests	Whether backup power works when called on	Test log with timestamps and load	Monthly
Lifts: AMC visits, certificate, entrapments, restoration time	Whether availability and the inspection duty were met	AMC reports, inspector certificate, breakdown log	Monthly
Landscaping: schedule completed, plant loss	Whether grounds were kept, not merely visited	Gardener schedule, site walk	Weekly, monthly
Waste: segregation at the bin, quantity against manifest	Whether waste left correctly segregated	Spot check, weighbridge or recycler receipts	Daily, monthly

Write the output measure for every family into the schedule as an obligation, and let the scorecard carry exceptions plus the families that are drifting: cleaning, pantry and HVAC on a tenant-facing site; on a campus, HVAC, DG, lifts and waste.

Severity: response, restoration and resolution are three clocks

A severity level is a statement about consequence, not about how inconvenient the fault is to fix. Once someone says “we couldn’t get the part”, they are describing resolution.

Response is acknowledgement and the arrival of someone competent. Restoration is the service usable again, even temporarily. Resolution is the root cause closed, with no recurrence inside an agreed window. A contract with one number per fault measures none of the three.

A severity table you can write

Level	Defined by	FM examples	Target shape
P1	Immediate loss of a critical environment, a safety exposure, or statutory compliance	Chiller or CRAC failure serving a server room; power failure where the DG will not take load; lift entrapment; a fire pump losing pressure	Response in minutes, restoration in hours, escalation above the supervisor at once
P2	Significant degradation, contained, with a workaround	Washroom out of consumables across a floor; one of two lifts out; an AHU down outside a critical area	Response within the shift, restoration same or next working day
P3	No immediate consequence: a defect, an asset, a missed task	A flickering light; a chipped tile; a deep clean deferred; a landscaping task pushed for weather	Response on the next visit, resolution within the month

The targets are yours: each level needs a target on all three clocks, in hours for P1 and days for P3, with a named escalation. A washroom out of consumables is a P2; a chiller failure in a data hall is a P1.

Service credits: size them to be recoverable, not punitive

A credit works only if the provider can pay it out of the contract margin and still have a reason to fix the problem.

CRISIL Ratings put the operating margin in the organised security and facility management segment at about 5%, against ₹1.15 lakh crore of revenue in FY2024 (CRISIL Ratings, 11 March 2025). On a ₹10 lakh monthly fee that is roughly ₹50,000, so a credit of 5% of the fee is the month’s whole profit. A schedule reaching a multiple of that is not a penalty; it is an unsecured claim, settled by negotiation rather than a bank transfer.

So step credits up from zero toward a ceiling the margin can carry, calculate on the affected service line rather than the whole fee, and cap the month.

What to leave out of the credit

• Access denial. Your own staff would not let the team in.
• Force majeure, including a shutdown ordered by the state.
• Buyer-caused delay. Spares in your approval queue, a permit not renewed.
• The first month of a mobilisation. Handover is covered under mobilisation and governance.
• Documented pre-existing condition. A chiller at the end of its life is a capital item.

Failure across three consecutive months belongs in a cure period, not an ever-larger deduction.

Inspection and sampling: a score filled in by the doer is not evidence

Inspection turns a completion log into evidence, and how you sample decides whether the score means anything. Sample by risk.

• Every washroom, every walkthrough. The failure surface users notice first, and the cheapest to inspect.
• Every critical asset, once a cycle. Chillers, AHUs, pumps, DG sets and panels. The rest of the plant can rotate.
• A rotating subset of rooms on each floor, fixed in advance so every area is seen within the month. Treat the count as a working reference, not an industry rule.
• A joint inspection monthly, unannounced checks between. The joint one settles the score with the supervisor present; the unannounced ones are why it stays truthful.

Independence is the part buyers skip. The principle in our field supervision benchmark applies unchanged: supervision must be independent of the post it verifies. If the site supervisor both cleans and scores, the buyer’s facilities lead or a field officer should hold the score. Misses become dated actions with a named owner. Completion, attendance and exceptions reported against the same roster, as on the operating visibility layer, make the trend checkable.

The monthly pack and the review that has to decide something

The monthly review exists to make four decisions, and the pack should fit on one page. At thirty slides nobody decides anything.

The one-page pack

1. Completion against the schedule, by service family, with missed items named.
2. Open exceptions, age in days, and how many closed on first visit.
3. The inspection score, areas inspected, and the misses in plain words.
4. The top complaints by area, and whether any repeat last month’s.
5. Open defects and statutory items, with the next due date.
6. Credits applied, calculated against the schedule rather than agreed in the room.
7. Decisions needed from the buyer, one line each.

Underneath sits the layer that stops the month arriving as a surprise. A daily exception call covers only what went wrong yesterday and who owns it. A weekly review takes aged exceptions, slippage against next week’s schedule and the consumable position. The reporting cadence we commit to is the same shape.

The room should be small enough to decide: your facilities owner, the provider’s account owner and supervisor, finance when a credit applies, and whoever owns the pantry users. The meeting has worked if it ends with one of four outcomes: the month accepted, the credit applied, a scope change approved, or the account on a performance notice.

Testing the scorecard before you sign it

Run the whole scorecard for one month on last month’s records, before signature. It is the cheapest test in the procurement and catches most schedules that would have failed in month four.

• Confirm every KPI has a data source that exists today. Name the system, register or person. A measure needing a system nobody bought will be estimated.
• Name who fills each line. Buyer or provider, by role. A line owned by nobody is the line filled in retrospectively.
• Check what “complete” means in month one. Definitions widen quietly between draft and first report: a task becomes complete when started, an area becomes cleaned when the bins were emptied. Put the acceptance standard in the contract.
• Run the credit formula on the dry-run month. Nothing on a plainly bad month means too small; consuming the whole margin means too large.
• Record the baseline defects with a photograph and a date.

The measures sit on top of a scope, and the scope design is where that work starts. The vendor selection guide covers the comparison, and the contractor audit method covers what to verify first.

Anti-gaming patterns to watch

• Completion logged by the doer. A tick with no inspection behind it is a claim, not a record.
• Complaints closed on the call and reopened later. Track the reopen rate, not closure alone.
• Inspection scores that only rise, in a building nobody refurbished. That is a scoring change.
• Headcount reconciled on paper. The roster matches the invoice and the shift does not. Check deployment at 8am and 11pm.
• Scope shrinking to match the score. A frequency drops from daily to alternate days, a deep clean becomes a wipe.
• Consumables counted at the store, not the dispenser. A full store and an empty dispenser is a distribution failure.

Frequently asked questions

How many KPIs should an FM contract have?

Few enough that each gets a decision in the monthly review: one output and one outcome per family you score, plus exception counts. Forty measures spread attention so thin that the misses nobody discusses matter most.

Is attendance a KPI?

It is a check on deployment, not a measure of service. Keep it as a gate: deployment matches the roster in headcount and shift, and a shortfall is adjusted on the invoice. Scored as a service KPI, it makes the provider optimise the roster while the building goes unmeasured.

What is a fair service credit?

One the provider can pay and still want the contract. With segment margins near 5%, a credit of 5% of the monthly fee consumes the whole month’s margin, so step up towards that ceiling rather than start there.

How often should the SLA be reviewed?

Daily and weekly in operations, formally once a month against a fixed one-page pack. At renewal, check whether the measures still describe the building you run.

Should every site use the same SLA?

Share the measures and definitions; set severity and targets per site. A chiller failure on a campus with a server room and one in a warehouse car park carry different consequences, and one severity table forces both to the pace of the less critical.

The bottom line

A facility SLA is a measurement design, and most of the value sits in three decisions: which lines are outputs rather than inputs, how severity is defined by consequence, and whether a credit is sized so the provider can pay it and carry on. Get those right and the review becomes a short meeting that decides something.

Send us the SLA and reporting pack you use today. We will read the measures against your building, tell you which lines will not survive a bad month, and show you where the pack and the floor have drifted apart.