Reactive maintenance is the most expensive kind. Most facilities still run on it.
Last updated
Zepth Edge module
30–40%+
the savings of a predictive maintenance programme over running assets to failure
US DOE Federal Energy Management Program, O&M Best Practices Guide
The same guidance puts predictive at 8–12% cheaper than a purely preventive, calendar-driven programme.
>55%
of facilities still run predominantly reactive maintenance — which is why the savings are still on the table
US DOE FEMP survey data
≥90%
PM and schedule compliance is the world-class benchmark — 95%+ on critical assets. Below 80%, the programme is not functioning
SMRP best-practice metrics
$1 → $4
every dollar of deferred maintenance escalating into roughly four dollars of future capital renewal
Biedenweg / Pacific Partners research
This is a maintenance-deferral multiplier, and it is not the 1-10-100 quality-cost rule used elsewhere on this site. Different research, different mechanism — deferral compounds into capital, rather than a defect getting more expensive to fix as it travels.
A work order is the unit of record for maintenance: the request, its priority, the planned work, the parts and labour consumed, and the evidence that closed it.
The system around that record — reactive fixes, preventive schedules, meter-based triggers, and the backlog sitting between them — is what decides whether an asset portfolio is being managed or merely repaired. The cost difference between those two things is documented, and it exceeds 30–40%.
The US Department of Energy’s own O&M guidance is unusually blunt about this. A predictive programme runs 8–12% cheaper than a calendar-driven preventive one, and more than 30–40% cheaper than running assets to failure. And its survey data finds that most facilities are still, predominantly, running assets to failure.
That gap has survived decades of everyone knowing about it, which tells you the problem is not knowledge. It is that reactive maintenance is self-sustaining: emergencies consume the budget, so the preventive work slips, so there are more emergencies. The cycle funds itself out of the money that would have broken it.
In a hotel the stakes are not merely financial, because the failure is in front of a guest. A PTAC that dies on a hot night is fifteen calls to the front desk, comped rooms, and reviews that name the broken air conditioning by name. The industry saw this at scale after 2020: deferred maintenance showed up directly, and visibly, in guest-satisfaction scores for facilities.
And deferral compounds. Research prices a dollar of deferred maintenance at roughly four dollars of future capital renewal — the repair you did not do becomes the replacement you cannot avoid, and it arrives as a capital request rather than an operating one.
The maintenance mix is a choice, and criticality should make it per asset. Reactive, calendar-based preventive, meter-based, condition-based. None of the four is wrong; running all your assets on one of them is. The corridor extract fan can reasonably run to failure. The chiller cannot. That is what the criticality classification on the asset register is FOR — it is not a taxonomy exercise, it is the input that decides how each asset gets maintained.
PM compliance is the integrity metric, and it is the one that gets quietly abandoned. 90% or better is world-class, 95%+ on critical assets. Below 80%, the schedule has become fiction — and the honest question is which of two things is true: you are understaffed, or you have generated more PMs than the asset base actually needs. Both are fixable. Neither gets fixed while the number is unmeasured. Measure it weekly, by trade.
Backlog is measured in crew-weeks, not in ticket counts. A count tells you nothing — two hundred lightbulbs and two hundred compressor rebuilds are the same number. The convention is 4–6 crew-weeks of total backlog, of which 2–4 weeks is “ready” work: planned, kitted, parts on the shelf, ready to hand to a technician. A backlog that is growing is the earliest warning you get of understaffing or PM over-generation, and it gives you that warning months before anything fails. Age-cap items at ninety days — an item nobody has done in three months is not a backlog item, it is a decision nobody has made.
Hotels have their own physics: rooms, occupancy windows, and OOO versus OOS. A quarterly deep-PM rotation per guest room is the common convention — coils, drains, sealants, hardware, safety devices — though brand standards vary and it is a convention rather than a standard. Noisy work belongs in occupancy-aware windows. And the out-of-order versus out-of-service distinction matters more than it sounds: out-of-order removes a room from available inventory, which flatters your occupancy percentage while costing you real revenue, and out-of-service keeps it sellable for a same-day fix. Misusing OOO to protect an occupancy number distorts RevPAR comparability, and it does so in the direction that makes you feel better.
Failure codes are what make maintenance history worth having. A problem-cause-remedy taxonomy on closure turns a pile of completed work orders into a reliability record you can run a Pareto analysis on: which assets fail, why, and what actually fixed it. Free-text closure notes destroy that. “Fixed it” is not data. Every hour of technician time is being spent either way — the only question is whether you get the reliability record as well as the repair.
Emergencies consume the budget. Preventive work slips because there is neither the money nor the crew. So there are more emergencies, which consume more of the budget. The cycle is not a failure of discipline; it is a stable equilibrium, and it will hold for years.
The guest-room PM gets skipped, so the coil fouls, so the unit fails on the hottest night of the year — which is precisely when it is least survivable, and when the comps and the reviews arrive together.
And with no cost captured against each work order, maintenance spend cannot be attributed to anything. So the budget is defended by anecdote at the annual review, against a finance team holding numbers. That is not an argument the maintenance manager wins, and it is not one they should have to have.
Meanwhile the deferred items accrue quietly, at roughly four dollars of future capital for every dollar not spent — until they arrive, all at once, as a capital shock that nobody forecast and everybody could have.
Work orders carry their cost — time and parts recorded against the order, so maintenance spend is attributable by asset, by trade and by category rather than defended by anecdote. PM schedules run on the calendar and on meters. Backlog and compliance are dashboards, in crew-weeks, by trade.
And every order is tied to its asset’s history, which is the point of the whole exercise: the chiller’s own file is what tells you it is time to stop repairing it. That is a maintenance record on the way in, and a capital business case on the way out.
Maintenance spend is attributable by asset and by trade — so the budget is defended with numbers rather than with anecdote.
PM compliance and backlog are measured, so understaffing and PM over-generation are caught months before anything fails.
The reliability record actually exists, because closure carries failure codes rather than free text.
Replace-versus-repair becomes a calculation, and the asset’s own history is the business case.
Time and parts recorded against the order and therefore against the asset — the step that makes every downstream number possible.
Calendar-driven where that fits, runtime- and meter-driven where it fits better. Criticality decides which.
Total and ready, by trade, with ninety-day age-capping — because a ticket count tells you nothing about the work.
Weekly, by trade, against the ≥90% benchmark. Below 80% is a signal, not a statistic.
Problem, cause, remedy on closure — which is what turns completed work orders into a reliability record.
Parts tied to the order and the asset, so “ready” work is genuinely ready and a planned job does not stall at the store.
From a guest, a staff member, an inspection finding, or a meter trigger. Four sources, one queue — because a fault reported by a guest and a fault predicted by a runtime meter are the same work.
Safety first, then guest-impacting, then asset-protecting, then cosmetic — weighted by how critical the asset is. This is where the register’s criticality classification earns its keep.
Scope, parts, permits, access, night windows. Planning is what converts a backlog item into “ready” work — and ready work is the only kind a technician can actually start.
Not to a cost centre, not to a month. To the order, and therefore to the asset. This is the step that makes every downstream number possible.
Problem, cause, remedy. Free text is a note; a code is a record. Only one of them can be analysed later, and later is when it matters.
Into PM optimisation, and into the replace-versus-repair decision. An asset with rising repair frequency and rising repair cost is telling you something, and the history is the only place it is written down.
Rising repair frequency and rising repair cost per asset, surfaced from meter and failure history — and handed to the capital process as a business case rather than an opinion. The chiller’s own file makes the argument.
By trade. A backlog trending up is the earliest warning of understaffing or PM over-generation available anywhere in the operation, and it arrives months before a failure does.
Suggested priority weighted by asset criticality and guest exposure — because the work order that will produce fifteen phone calls on a hot night does not announce itself as urgent when it is raised.
“Top ten assets by maintenance cost this year, with their open orders.” Answered from the work-order history, which is the question the budget conversation is actually made of.
The engineer’s judgment stays in charge; the AI removes the latency and the blind spots.
PM compliance weekly, by trade, against the ≥90% benchmark. Backlog in crew-weeks — total and ready — with ninety-day ageing. Maintenance cost by asset, by trade and by category, which is what turns the budget conversation from anecdote into arithmetic. Failure-code Pareto by asset class. And the replace-versus-repair candidates, with the history behind each one.
The US Department of Energy’s FEMP guidance puts a predictive programme at 8–12% cheaper than a calendar-driven preventive one, and more than 30–40% cheaper than running assets to failure. Its own surveys find most facilities are still predominantly reactive — which is precisely why those savings are still sitting on the table decades after they were first published.
Read the full answer90% or better, and 95%+ on critical assets. Below 80% the schedule has stopped describing reality, and the useful response is diagnostic rather than disciplinary: either the team is understaffed, or the PM programme has generated more work than the asset base actually needs. Both are fixable. Neither gets fixed while the number goes unmeasured.
The working convention is 4–6 crew-weeks of total backlog, of which 2–4 weeks is planned, kitted and ready to hand to a technician. Measure it in crew-weeks rather than ticket counts — two hundred lightbulbs and two hundred compressor rebuilds are the same count and nothing like the same problem — and age-cap items at ninety days.
Read the full answerA quarterly deep-PM rotation per room is the common convention — coil cleaning, drains, sealants, hardware, safety devices — coordinated floor by floor with the front office so the noisy work lands in occupancy-aware windows. It is worth being precise about this: quarterly is a convention rather than a standard, and brand standards vary.
Out-of-order removes the room from available inventory entirely: your occupancy percentage looks better, and you have lost roughly a night’s ADR in real revenue. Out-of-service keeps the room sellable, for brief same-day work. The trap is using OOO to protect an occupancy figure — it distorts RevPAR comparability, and it distorts it in the flattering direction, which is exactly why the habit persists.
Read the full answerProperty operations and maintenance in hotels runs at roughly 4–4.5% of total revenue, with labour accounting for about half of it — and that figure excludes utilities and capital expenditure. Older properties trend higher, which is the deferral multiplier arriving on the operating line before it arrives on the capital one.
Related answers
In the guides
Terms defined here
Zepth is the construction project delivery platform — it runs construction, procurement and asset management on one record, and does the work: reading the drawings, reviewing the submittals, matching the invoices and flagging the risks, with a human sign-off on anything consequential.
A short, tailored walkthrough on your real workflow — no generic demo.