Preventive vs Predictive Maintenance for Buildings

The quarterly service on the air handler was signed off in May. Filters changed, belts checked, bearings greased, box ticked. Then, on the hottest afternoon of July, the supply fan bearing seized and three floors lost air. The bearing had been getting louder for weeks. Nobody measured it, because the checklist asked whether the bearing had been greased, not whether it was failing.

That gap is the whole preventive vs predictive maintenance question. Preventive maintenance services equipment on a schedule. Predictive maintenance services it when a measurement says it needs it. Reactive maintenance waits for it to break. Most commercial buildings need all three, assigned asset by asset, yet much of the advice online is written for factories, not buildings.

This guide covers the definitions, predictive techniques for building equipment, the federal cost evidence, an asset-by-asset matrix, benchmarks, a roadmap and DMV specifics. If you want a team to run the program, our preventive maintenance service covers buildings across DC, Northern Virginia and Maryland.

Preventive vs predictive maintenance: what is the difference?

Preventive and predictive maintenance differ in what triggers the work. Preventive maintenance is done on a fixed calendar or runtime schedule whether or not the equipment needs it, while predictive maintenance is done when measured condition data, such as vibration, temperature or oil quality, shows that a specific component has started to degrade.

The Department of Energy’s Federal Energy Management Program (FEMP) defines preventive maintenance as actions “performed on a time- or machine-run-based schedule that detect, preclude, or mitigate degradation” of a component. It defines predictive maintenance as “measurements that detect the onset of system degradation,” so the causes can be removed or controlled before the component deteriorates significantly.

A building example makes it concrete. Changing air-handler filters whenever the calendar says so is preventive. Changing them when the pressure drop across the filter bank reaches the manufacturer’s limit is condition-based. Trending that pressure drop and scheduling the change before it reaches the limit is predictive.

Factor	Preventive	Predictive
Trigger	Calendar date or run hours	Measured condition or trend
Typical task	Change filters, grease bearings, adjust belts	Vibration route, infrared scan, oil sample, trend review
Best at catching	Wear-out failures with a known life	Failures that give measurable warning
Misses	Failures that develop between visits	Failures that give no warning
Up-front cost	Low: labor and parts	Higher: instruments, analysts, data
Main risk	Unneeded work, damage from intrusive work	False alarms, data nobody acts on
Best fit	Most building equipment	Critical, costly or long-lead assets

The two are not rivals. A predictive program sits on top of a preventive one, and preventive rounds are often where predictive readings get taken.

What are the main types of maintenance?

The main types of maintenance are reactive (including deliberate run-to-failure), corrective, preventive (time- or usage-based), condition-based, predictive and reliability-centered maintenance. They differ in what triggers the work: a breakdown, a found defect, a calendar date or meter reading, a measured condition, a forecast, or a structured analysis of how each asset fails.

Type	What it is	Trigger	Building example	Best fit
Reactive / run-to-failure	Fix after failure	Breakdown	Restroom exhaust fan replaced when it stops	Cheap, non-critical, redundant items
Corrective	Fix a known defect	Inspection finding or failure	Cracked belt found on a PM, replaced next day	Follow-up to any inspection
Preventive, time-based	Scheduled service	Calendar	Seasonal cooling-tower and boiler service	Wear items with a known life
Preventive, usage-based	Scheduled service	Run hours, starts or cycles	Generator service by engine hours	Equipment with a meter or BAS runtime
Condition-based	Service on measured condition	Reading crosses a limit	Filters changed on pressure drop	Wear you can measure simply
Predictive	Forecast failure from trends	Trend projects a limit crossing	Rising wear metals in chiller oil	Critical, costly, long-lead assets
Reliability-centered (RCM)	Pick a strategy per failure mode	Structured failure analysis	Pumps get vibration checks, small fans run to failure	Critical systems and portfolios
Proactive	Remove root causes	Repeat failure	Correcting the misalignment that keeps destroying bearings	Repeat offenders

Much of the preventive vs predictive vs corrective maintenance confusion comes from two terms. Planned is not the same as preventive: a well-planned repair after a failure is planned work, not preventive work (see our facility maintenance KPIs guide). Reliability-centered maintenance is a method for choosing among the other types, not a separate task. The Pacific Northwest National Laboratory (PNNL) calls it a strategic combination of reactive, preventive and predictive approaches, and SAE JA1011 sets the minimum criteria a process must meet to be called RCM.

What is reactive maintenance, and when is run-to-failure the right choice?

Reactive maintenance is repair or replacement after equipment fails, with no work done beforehand to prevent the failure. Run-to-failure is the deliberate version: a documented decision that an asset is cheap, non-critical or redundant enough that letting it fail costs less than maintaining it.

Unplanned reactive work is a default; run-to-failure is a strategy. That settles most reactive vs preventive maintenance debates: the question is not whether anything runs to failure, but whether anyone decided it should. FEMP calls reactive maintenance the “run it till it breaks” mode and lists its costs: unplanned downtime, overtime labor, larger parts inventories, secondary damage and shorter equipment life. Left unmanaged, it becomes the backlog described in our guide to deferred maintenance in commercial buildings.

NASA’s facilities guide to reliability-centered maintenance accepts reactive maintenance when it is “a conscious decision” that weighs the risk and cost of failure against the cost of preventing it. Its examples include restroom exhaust fans, small low-temperature water heaters, and non-critical electric motors under 7.5 horsepower.

Letting an asset fail is the right call when all of these are true:

1. Failure creates no safety, life-safety or code problem.
2. Failure causes no secondary damage. A small water heater qualifies only if a leak will not soak the ceiling below.
3. A replacement is easy to source locally, or a spare is on the shelf.
4. Occupants can tolerate the outage, or a redundant unit carries the load.
5. Maintaining the item would cost more over its life than replacing it.
6. The decision is written into the asset register, so nobody skips a required item by mistake.

Code-required systems are never candidates. Fire alarm and sprinkler systems, emergency lighting, elevators and backflow preventers have mandated inspection and testing regimes, covered in our fire and life safety inspection guide.

What is preventive maintenance?

Preventive maintenance is scheduled inspection, cleaning, lubrication, adjustment and parts replacement done at fixed intervals of time or use, regardless of the equipment’s current condition. In a commercial building it covers tasks like air-handler filter and belt service, boiler cleaning before the heating season, cooling-tower startup in spring, and roof and drain inspections.

Typical preventive scope:

• HVAC: filters, belts, coil cleaning, condensate drains, refrigerant checks, seasonal startup and shutdown. See our HVAC preventive maintenance checklist.
• Electrical: panel and switchgear inspection, exit and emergency lighting tests, torque checks by licensed electricians.
• Plumbing: water heaters, pumps, valves, drains, and required backflow testing by certified testers.
• Envelope: roofs, drains, sealants, doors and hardware.
• Life safety: the inspection, testing and maintenance that fire codes require, done by licensed contractors.

Our guide to building a preventive maintenance program covers the inventory, schedules and work orders behind this list.

Preventive maintenance has one structural weakness: it assumes failures are tied to age. PNNL, summarizing airline, Department of Defense and Navy studies, reports that age-related failures account for less than 20 percent of failures and random failures for roughly 80 percent. Those studies came from aviation, defense and other industrial equipment, not commercial buildings, but the point carries: a fixed interval cannot prevent a random failure. NASA adds that imposing arbitrary preventive tasks, such as taking working equipment apart to replace parts, can raise failure rates by resetting early-life failures.

The practical fix is to make preventive visits measure, not just service. A technician who records amps, temperatures, pressures and filter pressure drop on every visit is building the data that condition-based triggers need. That is a low-cost first step toward predictive maintenance.

What is predictive maintenance and how does it work in a commercial building?

Predictive maintenance uses periodic or continuous measurements of equipment condition to detect degradation early and schedule the repair before failure. In commercial buildings the main techniques are vibration analysis on rotating equipment, infrared thermography on electrical gear, oil analysis on chillers and generators, ultrasound for leaks and bearings, and trend data from the building automation system.

The idea behind it is the P-F curve: a fault becomes detectable at a point of potential failure (P) before it ends in functional failure (F). The goal is to detect it early enough to plan, order parts and schedule the work. For vibration, NASA’s guide says narrowband analysis can warn of impending failure “several weeks or months in advance.”

Technique	What it finds	Building assets	Who does it
Vibration analysis	Imbalance, misalignment, looseness, bearing and belt flaws	Chillers, large pumps, fans, cooling-tower gearboxes	Trained analyst, often a contracted route
Infrared thermography	Hot connections, overloaded phases, blocked flow, roof moisture	Switchgear, panels, motor control centers, roofs	Qualified thermographer; licensed electrician corrects
Oil and wear-particle analysis	Lubricant breakdown, contamination, wear metals	Chiller compressors, generator engines, gearboxes	Technician samples, lab analyzes
Ultrasound	Leaks, electrical discharge, dry bearings	Compressed air, steam traps, valves, electrical gear	Technician with a detector
Motor analysis	Winding and insulation faults	Large motors	Electrical tester
BAS and sensor trends	Drift from normal operation	Air handlers, rooftop units, chillers, boilers, pumps	Operator or analytics software

NASA’s guide uses a chilled water system as its worked example. Its list includes flow, temperature difference (which reveals fouled tubes), pressure difference across pumps and chiller, motor power, vibration, lubricant analysis, and thermography of the motor control centers.

How alerts work. NASA describes setting an alert at a statistically significant deviation from the mean, usually two standard deviations, and an alarm at three. Once a reading passes the alert, the guide says to cut the monitoring interval to one-third to one-quarter of the previous one, so a pump checked quarterly moves to every three to four weeks.

The building automation system is already a sensor network. Supply air temperature, static pressure, valve positions and runtime all reveal equipment drift. PNNL’s Building Re-tuning process uses BAS data to find operational problems and fix them at low or no cost. In re-tuning work with federal agencies, PNNL reports building energy savings typically between 5 and 25 percent. Many faults that trend data exposes, such as stuck dampers and valves that leak through, are maintenance faults.

Every alert needs a work order and a due date, the flow in our guide to the work order management process.

Is predictive maintenance the same as condition-based maintenance?

Predictive and condition-based maintenance overlap so much that many sources, including the Department of Energy’s FEMP guide, treat the terms as interchangeable. The stricter definition, in the European maintenance terminology standard EN 13306, makes predictive maintenance a type of condition-based maintenance that adds a forecast of how the degradation will progress.

In practice, condition-based work acts when a reading crosses a limit: the filter pressure drop hits its limit, so the filters get changed. Predictive maintenance uses the trend to say when the limit will be crossed: pump bearing vibration is rising steadily, so the replacement goes into the next planned shutdown.

Under EN 13306, both sit inside the preventive branch, alongside predetermined (calendar or runtime) maintenance. That framing is more useful than the “preventive vs predictive” headline, because every approach except reactive work aims to act before failure. For a building owner the label matters less than three questions: is there a measurement, is there a limit, and does crossing it create a work order?

Is predictive maintenance worth the cost?

Predictive maintenance is worth the cost on assets where failure is expensive, disruptive or slow to fix and where the failure gives measurable warning. Federal guidance cites meaningful savings over preventive programs, but those figures come from older, largely industrial studies, so a building should justify predictive work asset by asset, not as a blanket program.

FEMP’s Operations and Maintenance Best Practices Guide, Release 3.0 (2010) cites studies estimating that preventive maintenance saves 12 to 18 percent over a purely reactive program, and that a properly functioning predictive program saves 8 to 12 percent over preventive maintenance alone. It also cites independent surveys of industrial predictive programs reporting a tenfold return on investment, a 25 to 30 percent reduction in maintenance costs, a 70 to 75 percent elimination of breakdowns, and a 35 to 45 percent reduction in downtime. PNNL repeats the 12-to-18 and 8-to-12 percent figures in its current maintenance approaches best practice.

Read those numbers with care. The guide dates from 2010 and does not identify the underlying studies, the survey figures are explicitly industrial, and none is specific to office or multifamily buildings. They show direction, not a forecast. FEMP itself lists the downsides: investment in diagnostic equipment, investment in staff training, and “savings potential not readily seen by management.”

The real cost drivers for a building:

• Buying versus contracting. A contracted infrared or vibration route avoids owning instruments and employing an analyst.
• Analyst skill. NASA lists the analyst’s knowledge and experience among the factors that decide whether vibration monitoring works.
• False alarms. NASA cites Navy studies of vibration monitoring with detection probabilities between 0.76 and 0.92 and a false-alarm rate of 0.08. Every false alarm costs a site visit.
• The repair itself. Predictive maintenance does not remove the repair. It moves it to a better time, at planned labor rates, with parts ordered in advance.

Predictive work pays where failure is costly and detectable, the warning period is long enough to act on, and there is no redundant unit.

Which building assets suit preventive, predictive or run-to-failure?

Most building assets need preventive maintenance as the base, with predictive techniques added on the few critical, expensive rotating and electrical assets, and run to failure reserved for cheap, non-critical, easily replaced items. Code-required systems such as fire protection, elevators and backflow preventers follow their mandated inspection and testing schedules regardless of strategy.

Asset	Base strategy	Predictive or condition layer	Run to failure?
Chillers	PM per manufacturer	Oil analysis, vibration, refrigerant and temperature trends	No
Cooling towers	PM plus water treatment	Fan and gearbox vibration, water chemistry trends	No
Boilers	PM plus required inspection	Combustion analysis, stack temperature trends	No
Air handlers	PM: filters, belts, coils	Filter pressure drop, fan vibration on large units, BAS trends	No
Rooftop units	Seasonal PM	BAS or thermostat alarms, refrigerant log	No; replace by plan
Large or critical pumps	PM	Vibration, motor current	No
Small redundant pumps	Light PM	Usually none	Often yes
Switchgear and panels	PM by licensed electrician	Infrared thermography	No
Emergency generator	Required testing plus PM	Oil, fuel and coolant analysis	No
Elevators	Contractor maintenance plus required inspections	Controller fault logs	No
Fire alarm and sprinklers	Code-required testing	None	Never
Backflow preventers	Code-required testing	None	Never
Roofs	Inspections, plus after storms	Infrared moisture survey	No
LED interior lighting	Spot replacement	None	Yes, except emergency lighting
Small exhaust fans	None	None	Yes
Small water heaters	Basic checks	Leak sensor	Often, with leak protection

Electrical, elevator, fire protection and backflow work belongs to licensed or certified professionals. In buildings Levaru manages, routine work is self-performed and those trades are handled by licensed subcontractors under Levaru’s supervision. For the electrical layer, see our commercial electrical maintenance guide.

What is a good planned vs reactive maintenance ratio?

A commonly cited rule of thumb is 80 percent planned work to 20 percent reactive, but it is an industry convention, not a standard. Federal benchmarks published by PNNL put typical programs at 40 to 60 percent reactive, set a target goal of 20 to 30 percent, and describe less than 10 percent reactive as best in class.

PNNL’s benchmark table, adapted from FEMP:

Approach	Typical range	Target goal	Best in class
Reactive	40–60%	20–30%	Under 10%
Preventive	30–50%	25–35%	25–35%
Predictive	10–15%	30–40%	45–55%

PNNL adds that achievable targets depend on each site’s mission and systems. A small office building with packaged rooftop units should not chase a best-in-class predictive share.

Three rules keep the ratio honest:

1. Measure labor hours, not ticket counts. Fifty light-bulb tickets and one chiller teardown are not comparable.
2. Type every work order when it is created, so emergencies cannot be relabeled later.
3. Report tenant service requests separately. A stuck door or a dripping faucet is reactive by nature, not a program failure, and an apartment building will usually log more of them than a warehouse.

Compare the building with itself over time.

How do you move from reactive to preventive to predictive maintenance?

Moving from reactive to predictive maintenance is a staged process: inventory and rank assets, put the critical ones on documented preventive schedules, record readings during every visit, switch tasks to condition-based triggers where the data supports it, then add predictive techniques where failure is costly. Each stage depends on the records the previous one produced.

1. Build the asset register. Record make, model, serial number, location, area served and install date for every maintainable asset, and tag each one. Our QR code asset tagging guide shows how.
2. Rank criticality by consequence. A simple building scale: life safety and code, tenant operations, comfort, and discretionary.
3. Baseline your current mix. PNNL suggests estimating the share of maintenance time or cost in each approach, and says a well-organized program can estimate within 5 or 10 percent.
4. Calendar the code-required work first. It is not a strategy choice.
5. Write preventive tasks from the manuals and failure history. Mark run-to-failure items explicitly.
6. Make every visit capture readings: amps, temperatures, pressures, filter pressure drop and photos.
7. Switch to condition-based triggers where the readings support it, starting with filters and belts.
8. Pilot predictive work on the top tier. PNNL advises starting with a pilot on a problem system. For most buildings that means an infrared scan of the electrical distribution, oil analysis on chillers and generators, and a vibration route on the largest rotating equipment.
9. Use the BAS. Set alarms on the trends that matter.
10. Find the root cause of repeat failures. A bearing that keeps failing points to misalignment, lubrication or installation. That is proactive maintenance.
11. Review and adjust. Track the reactive share, PM findings and repeat failures, then change intervals based on the records.

None of this works without a system of record. In buildings Levaru manages, owners can see PM schedules, inspection results (pass/fail) and work orders with photos in the client portal of the CMMS included with the service, and can export the asset register and work-order history.

What changes for buildings in DC, Virginia and Maryland?

In DC, Virginia and Maryland, the strategy choice is shaped by code-required maintenance and by energy performance laws. Virginia’s Maintenance Code requires building systems to be kept in good repair, and the DC and Maryland energy performance standards turn quiet HVAC efficiency drift into a compliance risk, which strengthens the case for condition monitoring.

Virginia. The Virginia Maintenance Code (Part III of the Uniform Statewide Building Code) states in Section 103.2 that “buildings, structures, and systems shall be maintained and kept in good repair.” Section 103.2.1 lets nonrequired components and systems be discontinued “provided that no hazard results,” and Section 103.2.3 puts the duty on the owner unless it is expressly or impliedly the tenant’s. Read together, those rules leave room to retire nonrequired items if no hazard results, but not required systems. Section 103.2.2 adds an exception: nonrequired fire protection systems must still be maintained to function as originally installed.

District of Columbia. DC’s Building Energy Performance Standards come from the Clean Energy DC Omnibus Act of 2018 and are administered by the Department of Energy and Environment. Covered buildings that fall short of their standard must follow a compliance pathway. Our BEPS deadlines guide covers the dates.

Maryland. Under the state’s Building Energy Performance Standards, owners of most commercial and multifamily buildings of 35,000 square feet or larger must submit an annual benchmarking report and may need improvements to meet performance standards in the 2030s. Montgomery County runs its own standard, which covers most buildings of 25,000 square feet and larger.

Why this matters: fouled coils, failed economizers, leaking valves and simultaneous heating and cooling rarely cause a breakdown. They raise energy use month after month, and BAS trend monitoring catches them better than a checklist does. In a covered building, that makes condition monitoring a compliance tool, not only an energy one. The region’s humid summers and cold winters also load equipment from both directions, so condensate drains, coils and freeze protection deserve condition checks.

This guide is general information, not legal or engineering advice. Confirm requirements with your jurisdiction and with licensed professionals.

Frequently asked questions

What is the difference between preventive and corrective maintenance?

Preventive maintenance is scheduled work done before failure to reduce the chance of failure, while corrective maintenance fixes a defect after it has been found, by an inspection or a breakdown. Good preventive programs generate corrective work: the belt found cracked on a PM visit becomes a corrective work order, which usually costs less than an emergency call.

Is it preventive or preventative maintenance?

Preventive and preventative maintenance are both correct spellings and mean the same thing. Merriam-Webster says there is no difference between the two words and notes that preventive is used much more often; federal guidance such as the FEMP and NASA maintenance guides and standards such as NFPA 70B also use preventive maintenance. Pick one spelling and use it consistently.

Is predictive maintenance AI?

Predictive maintenance can use artificial intelligence and machine learning, but most building-scale predictive work relies on baselines, fixed limits and trend lines that a trained technician or a building automation system can apply. NASA’s facilities guide, for example, sets alerts at a statistically significant deviation from the mean, usually two standard deviations. AI helps most where there are many sensors and complex patterns.

What is proactive maintenance, and how does it relate to RCM?

Proactive maintenance removes the root causes of failure instead of repeatedly repairing the symptoms, using tools that NASA’s facilities guide lists as root-cause failure analysis, precision installation, commissioning and better equipment specifications. Reliability-centered maintenance is the method that decides, failure mode by failure mode, which approach each asset gets. Buildings usually need only a streamlined version.

What does NFPA 70B require for electrical equipment maintenance?

NFPA 70B, the Standard for Electrical Equipment Maintenance, moved from recommended practice to standard with its 2023 edition, which requires a documented electrical maintenance program. Industry summaries of its interval table, which applies where the manufacturer gives no interval, list infrared thermography every 12 months, or every 6 months for equipment in the poorest condition rating. As of October 2026 the 2026 edition is current, so confirm intervals with a licensed electrical contractor.

Does preventive maintenance need to be logged?

Preventive maintenance should always be logged with the date, the person who did the work, readings, findings and photos, because warranties, insurers, code officials and standards such as NFPA 70B rely on proof that maintenance happened. Logged readings are also the raw data predictive maintenance needs. A CMMS keeps that history attached to each asset.

Can a small building do predictive maintenance?

A small or mid-size building can do practical predictive maintenance without buying instruments or sensors. Contract an annual infrared scan of the electrical distribution, send oil samples from a chiller or generator to a lab, set BAS alarms on the trends that matter, and record readings on every preventive visit. Full sensor networks rarely pay off at this scale.