Predictive vs Preventive Maintenance: What's the Difference?

Predictive vs preventive maintenance: the short answer

Preventive maintenance services equipment on a fixed schedule — every 500 hours, every three months — regardless of its actual condition. Predictive maintenance uses sensor data (vibration, oil, temperature) to service equipment only when readings show it's starting to fail. Preventive is cheaper and simpler to start; predictive catches faults preventive schedules miss, but needs data and someone to act on it.

What is preventive maintenance?

Preventive maintenance is time-based or usage-based servicing: you change the oil every 500 hours, replace a belt every 12 months, inspect a forklift every shift — on a calendar or a meter, not in response to a measured symptom. It assumes that most components wear out in a roughly predictable way, so servicing them before that point is cheaper than waiting for them to fail.

That's the model most small manufacturers already run, and it works well for a lot of plant. We've covered how to set the intervals for common equipment — forklifts, compressors, pressure vessels — in our guide to building a preventive maintenance schedule. This article picks up the next question: when does a fixed schedule stop being enough?

Its honest weakness is that the calendar doesn't know the machine's real condition. You'll sometimes replace a part with life left in it (wasted cost), and sometimes miss a fault that develops between services (unplanned downtime) — because the interval is built on averages, not on what's actually happening inside that machine right now.

What is predictive maintenance?

Predictive maintenance (sometimes called condition-based maintenance, though the two aren't identical) monitors the actual condition of equipment while it runs, and triggers a service only when the data shows early signs of a developing fault — not on a fixed date.

The common data sources:

• Vibration analysis — accelerometers detect bearing wear, misalignment or imbalance in rotating equipment (motors, pumps, fans) well before it's audible or visible.
• Oil analysis — lab or on-site testing of lubricant for metal particles, contamination or chemical breakdown, which flags internal wear inside gearboxes, engines and hydraulics.
• Thermal imaging — infrared cameras spot abnormal heat in electrical connections, motors and bearings, often the earliest visible sign of a failing component.
• Ultrasonic and acoustic monitoring — picks up the high-frequency sound of air/gas leaks, electrical arcing or early bearing failure.
• Process sensors — pressure, flow, current draw and speed sensors flag equipment drifting outside its normal operating envelope.

The output of any of these is a trend, not a pass/fail. Someone (a technician, a CMMS, or increasingly software with a threshold or a model behind it) has to interpret that trend and decide when to act. Predictive maintenance is a data and decision-making system, not just a sensor purchase — that's the part vendors selling the sensors don't always lead with.

Predictive vs preventive maintenance: what's the difference?

	Preventive maintenance	Predictive maintenance
Trigger	Fixed schedule (time or usage)	Measured condition data
Typical cost to start	Low — a calendar, checklist or basic CMMS	Higher — sensors, monitoring software, often a service contract
Ongoing cost	Predictable labour and parts cost	Sensor/software subscription plus analysis time
Data or skills needed	Manufacturer's manual, a register	Someone who can read vibration/oil/thermal data, or a platform that does
Catches faults between services?	No — only what the schedule happens to catch	Yes, in principle — that's the whole point
Risk of unnecessary servicing	Higher — parts replaced on schedule whether needed or not	Lower — parts serviced closer to actual wear
Best suited to	Most general plant, low-criticality equipment	High-criticality, high-cost-of-failure, or hard-to-access equipment
Typical adopter	Small-to-mid manufacturers, most equipment	Larger sites, or specific critical assets within a smaller site

Neither replaces the other outright. Most sites that run predictive maintenance well are still running preventive maintenance on everything that doesn't justify the extra investment — predictive is usually layered onto a small subset of critical machines, not swapped in wholesale.

What does the data actually say about the payoff?

Vendor marketing tends to run ahead of the evidence here, so it's worth being precise about what's actually been measured — and upfront that the numbers vary a lot by source and age.

The most-quoted figures trace back to the US Department of Energy's Federal Energy Management Program Operations & Maintenance Best Practices Guide. Its benchmark put annual operating cost at roughly $18/hp/year for reactive maintenance, $13/hp/year for preventive, and $9/hp/year for predictive — implying predictive running 8–12% cheaper than preventive alone, and preventive 12–18% cheaper than reactive. The same source cited industrial averages of a 10x ROI and 35–45% less downtime for well-implemented predictive programs — but that benchmark is now well over a decade old and describes best-case results, not a typical one.

More recent, independent data is more modest. A 2018 PwC study of 268 European manufacturing and infrastructure companies found predictive maintenance cut maintenance costs by around 12% and improved equipment availability by around 9%, compared with each company's existing practice — real, but far smaller than the older DOE figures imply. Plant Engineering's industry surveys similarly show most manufacturers still run preventive maintenance as their core program, with predictive/condition monitoring adopted on a growing but partial basis, concentrated at larger sites and on specific critical equipment.

The honest summary: predictive maintenance can genuinely reduce cost and downtime, but the size of the benefit is contested and implementation-dependent. The more dramatic numbers circulating online trace back to a 2010-era benchmark, not a current, independently verified average — treat any vendor quoting a precise ROI figure with the same scepticism.

When is predictive maintenance worth it for a small manufacturer?

Predictive maintenance earns its cost when most of these are true for a specific asset:

• Failure is expensive — the machine is a bottleneck that stops the whole line, not just one station. See the real cost of unplanned downtime.
• Failure develops gradually and detectably — bearing wear, gearbox degradation and insulation breakdown show up in vibration, oil or thermal data well before a stoppage; sudden failures with no lead-up don't benefit from monitoring.
• The equipment runs enough hours to justify the sensor cost — a single-shift, lightly used asset rarely earns it back.
• Someone will act on the data — a trend nobody looks at is money spent for nothing.
• It's hard or costly to inspect manually — elevated, enclosed or inaccessible equipment benefits more from continuous sensing than equipment a technician can easily check by hand.

When a good preventive schedule is enough

For most equipment in most small-to-mid manufacturers, a disciplined preventive schedule — right intervals, actually followed, actually recorded — closes most of the gap at a fraction of the cost. That's true when:

• The equipment is low-cost to replace or repair, or a spare is on hand
• A stoppage is annoying rather than catastrophic to production
• The failure mode is genuinely time/usage-based (a filter that clogs at a predictable rate, a belt that wears at a predictable rate) rather than condition-variable
• You don't yet have anyone in the business who can interpret sensor data, and can't yet justify hiring or contracting that skill

If you're not confident your current preventive schedule is even being followed reliably, that's the more urgent problem to fix first — predictive maintenance on top of an undisciplined preventive program just adds cost on an existing gap. Our guide on whether you need a CMMS is a reasonable next step before adding condition monitoring.

FAQ

Is predictive maintenance the same as condition-based maintenance? They're closely related and often used interchangeably. Strictly, condition-based maintenance acts once a measured value crosses a set threshold; predictive maintenance uses trend data to forecast failure before that point. Most small manufacturers don't need to worry about the distinction in practice.

Do I need IoT sensors and AI software to do predictive maintenance? No. Basic predictive maintenance can start with a handheld vibration meter, periodic oil sampling sent to a lab, and a thermal camera on a walk-around — all manual, all far cheaper than a permanent sensor network. Continuous, connected monitoring is the expensive end of predictive maintenance, not a requirement for any of it.

Can a small manufacturer run both preventive and predictive maintenance at once? Yes — that's the normal setup. Most sites keep preventive schedules for the bulk of their equipment and add predictive monitoring only to the specific machines where failure is expensive and detectable early.

What's the first piece of equipment to try predictive maintenance on? Whichever asset would hurt the most if it failed unexpectedly and runs enough hours to justify the attention — commonly a critical motor, compressor, or gearbox on the main line. Starting with one asset on manual monitoring is a low-risk way to test whether the data is actually useful before spending more.

Does predictive maintenance definitely reduce costs compared to preventive maintenance? Usually, but the size of the saving varies by source and implementation — older benchmarks suggest a bigger gain than recent independent studies do. It also depends on whether the failure mode shows up in sensor data early enough to act on, and whether anyone is actually reviewing the readings.

Ardmor works with Victorian manufacturers on both sides of this — building maintenance schedules that actually get followed, and helping decide where the extra cost of condition monitoring is genuinely worth it rather than a vendor's sales pitch. If you want an outside view on where your maintenance spend is actually going and what it should be doing instead, book a lean/process and safety audit.

This article is general information for manufacturers, not professional advice. Last updated: 20 September 2026.