Preventive vs Predictive Maintenance: Which Should You Use?

In the preventive vs predictive maintenance debate, preventive services assets on a fixed schedule while predictive watches their actual condition and intervenes only when data signals failure. Neither wins outright. This guide compares reactive, preventive, predictive and prescriptive maintenance, shows where each pays off, and explains how to build a mixed strategy driven by asset criticality rather than dogma.

Preventive vs Predictive Maintenance: The Four Strategies at a Glance

The whole preventive vs predictive maintenance question sits on a single spectrum: from doing nothing until an asset breaks to predicting the failure weeks out and acting just in time. Four named strategies anchor that spectrum, and the EN 13306 maintenance terminology standard formalises most of them.

• Reactive (run-to-failure) - you fix the asset after it fails. Sometimes a deliberate, correct choice; often an accident.
• Preventive (PM, time- or usage-based) - you service the asset on a fixed calendar or meter interval, regardless of its actual condition.
• Predictive (PdM / condition-based, CBM) - you monitor a condition indicator (vibration, temperature, oil debris, current signature) and intervene only when it crosses a threshold that signals an emerging failure.
• Prescriptive (RxM) - analytics not only detect the developing fault but recommend the specific corrective action and its optimal timing, weighing production schedule and spares availability.

The honest answer to "preventive vs predictive" is that a healthy plant runs all four at once, assigned asset by asset. The skill is matching strategy to consequence and failure behaviour, not crowning a single winner.

Reactive Maintenance: The Default You Should Choose Deliberately

Run-to-failure means no scheduled intervention. You operate the asset until it stops, then repair or replace it. It carries zero planning overhead and extracts the full useful life from a component.

That makes it the right strategy for low-criticality, non-safety assets where failure is cheap, obvious and quick to fix: a $40 light fixture, a redundant pump in a parallel pair, a standby fan.

The trap is unplanned reactive work on assets that matter. An unplanned failure typically costs three to nine times a planned repair once you add collateral damage, expedited parts, overtime, secondary failures and lost throughput.

Use the Downtime Cost Calculator to put a dollar figure on an hour of unplanned stoppage on a given line. That number is the budget that justifies moving the asset onto a preventive or predictive plan.

When reactive is legitimate

• Consequence of failure is low and contained (no safety, environmental or major production impact).
• The asset has a random failure pattern that inspection cannot predict.
• Redundancy exists, so a single failure does not stop the process.
• Repair is fast and spares are on the shelf.

Preventive Maintenance: Trading Some Life for Predictability

Preventive maintenance (PM) services or replaces components on a fixed schedule - every 500 running hours, every quarter, every 10,000 cycles - before the statistical onset of wear-out.

Done well, it converts unpredictable breakdowns into planned, off-shift work with parts and labour staged in advance. It is the backbone of most maintenance programs and the foundation of TPM (Total Productive Maintenance, JIPM).

Where preventive maintenance shines

• Age-related (wear-out) failure modes - belts, filters, seals, lubricants, brake pads - where failure probability genuinely rises with time or usage.
• Statutory and safety tasks - pressure-vessel inspections, lifting-gear checks, calibration - where the interval is mandated regardless of condition.
• Low-cost consumables where inspecting to assess condition costs more than simply replacing the part.

The hidden cost: over-maintenance

A landmark finding behind Reliability-Centered Maintenance (codified in SAE JA1011) is that only a minority of failure modes are age-related. Nowlan and Heap's United Airlines study found roughly 11% of components showed a clear wear-out pattern that a fixed interval could address.

The remaining ~89% had random or infant-mortality patterns where a time-based PM does little - and can introduce failures through intrusive work, reassembly error and infant mortality. Over-serviced assets waste labour, consume parts and create their own breakdowns. This is precisely the gap predictive maintenance was created to close.

If you are building or rationalising PM tasks, the PM Procedure Generator produces standards-based procedures (EN 13306/13460, ISO 55001, OSHA 1910.147 LOTO) with measurable acceptance criteria, so each task earns its place rather than being copied from a manual.

Predictive Maintenance (PdM): Acting on Condition, Not the Calendar

Predictive maintenance - the condition-based maintenance (CBM) family in EN 13306 - monitors a physical parameter that degrades measurably before functional failure, then triggers work only when that parameter crosses an alarm threshold.

The core idea is the P-F interval: the window between the point a defect becomes detectable (P) and the point of functional failure (F). PdM aims to catch the fault early in that window and schedule a planned intervention before F.

Common PdM techniques

• Vibration analysis - the workhorse for rotating equipment (bearings, gears, imbalance, misalignment); ISO 20816 governs vibration evaluation.
• Infrared thermography - electrical connections, switchgear, overloaded motors, steam traps.
• Oil and wear-debris analysis (tribology) - gearboxes, hydraulics, large bearings.
• Ultrasound - early bearing faults, compressed-air and steam leaks, electrical arcing.
• Motor current signature analysis (MCSA) - rotor bars, eccentricity, load faults.

Pros and cons

PdM extracts nearly the full useful life of a component while still warning you before failure, slashing both over-maintenance and unplanned downtime. DOE and industry estimates commonly cite 8-12% cost reduction over preventive programs, plus substantial breakdown reductions.

The cost is real: sensors and instruments, analyst skill (or AI models), and a baseline data-collection routine. PdM also only works where a failure mode has a detectable, sufficiently long P-F interval - it cannot help with truly instantaneous failures.

Track whether it is working with MTBF / MTTR: a rising Mean Time Between Failures and a falling unplanned MTTR are the signatures of an effective PdM program.

Prescriptive Maintenance: From Diagnosis to Decision

Prescriptive maintenance (RxM) is the analytics layer on top of PdM. Where predictive tells you a bearing will fail in roughly three weeks, prescriptive tells you run at reduced load until Saturday's planned outage, when the part and crew are available, and order the spare now.

It fuses condition data, failure models, production schedules and inventory to recommend - and sometimes automate - the optimal action and timing.

RxM depends on a mature data foundation: reliable sensors, historised condition data, a connected CMMS/EAM, and often machine-learning models trained on labelled failure history. It delivers the highest theoretical value but also the highest implementation cost and complexity.

So it is reserved for the most critical or expensive assets, where a few hours of optimised timing is worth real money. Most plants reach prescriptive on a handful of assets long before it makes sense fleet-wide.

Preventive vs Predictive: The Comparison Table

The trade-offs become clearest side by side. "Best for" assumes the failure mode is compatible with each approach.

The Cost Curve: Why the Cheapest Strategy Is Rarely the Cheapest

Plot total cost (maintenance spend plus the cost of failures) against maintenance intensity and you get a U-shaped curve.

• Too little maintenance (pure reactive on critical assets) is expensive because failures dominate.
• Too much (over-frequent PM on everything) is expensive because labour, parts and intrusive-work failures dominate.

The minimum sits where you spend just enough to prevent the failures that actually hurt. PdM lets you sit closer to that minimum because you intervene on evidence, not on a conservative calendar guess.

The right benchmark for total spend is maintenance cost as a percentage of Replacement Asset Value (RAV). SMRP and industry sources place world-class around 2-3% of RAV annually; consistently above ~5% usually signals a reactive culture or over-maintenance.

Check where you sit with the Maintenance Cost % of RAV Calculator. To justify investing in condition monitoring or a CMMS, model the savings against avoided downtime with the CMMS ROI Calculator - a CMMS is usually the prerequisite that makes PdM data actionable.

Criticality First: How to Decide Asset by Asset

You cannot afford predictive maintenance on everything, and you should not run-to-failure on anything that can hurt people or the plant.

The decision is driven by criticality - the product of how likely a failure is and how bad its consequences are. This is the foundation of EN 15341 KPI thinking and ISO 55001 risk-based asset management.

A practical decision path

• Rank every asset by criticality. Score likelihood x consequence (safety, environment, production, cost) on a matrix. The Asset Criticality Calculator turns this into a defensible ranking.
• Understand the failure pattern. Use failure history and the Weibull Reliability Calculator to test whether failures are age-related (shape beta > 1, so PM works) or random/infant-mortality (beta around or below 1, so a fixed interval is largely wasted).
• Assign a strategy using a simple rule of thumb (below).

This is exactly the logic Reliability-Centered Maintenance (SAE JA1011/JA1012) formalises: for each failure mode, ask whether a condition task, a scheduled-restoration task, a scheduled-discard task, or run-to-failure is both technically feasible and worth doing.

The Role of IoT and Condition Monitoring

For decades, PdM meant a technician walking a route monthly with a handheld vibration meter. Cheap wireless sensors, edge computing and the Industrial IoT have changed the economics: continuous monitoring once reserved for a turbine can now ride on a mid-criticality pump.

That matters because the P-F interval sets your minimum sampling rate. If a bearing fault develops over two weeks, monthly readings will miss it - but a wireless vibration node sampling daily will catch it with room to plan.

IoT does not change the strategy logic. Criticality and failure mode still decide whether to monitor; IoT simply lowers the cost of monitoring, pushing the break-even point so more assets qualify for PdM.

The risks are real too: alarm floods, false positives that erode trust, and data with no analyst or model behind it. A sensor that no one acts on is pure cost. Start with the highest-criticality assets, prove the P-F detection works, then expand.

Building a Mixed Strategy That Actually Works

The mature answer to "preventive vs predictive" is a deliberately blended program, reviewed continuously. A pragmatic rollout:

• 1. Get the foundation right. A functioning CMMS/EAM with accurate asset registers, work history and failure coding. Without failure data you cannot tell age-related modes from random ones, and PdM has nothing to learn from.
• 2. Rank by criticality and concentrate effort where consequence is highest (criticality matrix).
• 3. Assign strategies by failure mode using RCM logic - PdM where a P-F interval exists, optimised PM for true wear-out, run-to-failure for the trivial.
• 4. Rationalise the PM program. Kill calendar tasks aimed at random failures; tighten or relax intervals using real reliability data. Regenerate clean, measurable procedures with the PM Procedure Generator.
• 5. Layer in condition monitoring on high-criticality assets, starting with vibration and thermography where the ROI is clearest.
• 6. Measure and iterate. Track MTBF/MTTR, planned-vs-reactive ratio (target 80%+ planned), backlog weeks (healthy 2-4 weeks), cost % of RAV and the throughput payoff via OEE. Use EN 15341 as your KPI framework so the numbers stay comparable over time.

Browse the full toolkit on the AMAADOR tools page. The goal is not to be "a predictive plant" - it is to spend the next maintenance dollar where it removes the most risk, which almost always means a portfolio of all four strategies, rebalanced as your data improves.

Related free calculators

Frequently asked questions