Preventive, Predictive, or Condition-Based Maintenance: Which Does Each Asset in Your Plant Need?
The Question Is Wrong
When a vendor asks, “Do you use preventive or predictive maintenance?”, the honest answer from almost any plant should be: both, plus corrective maintenance—and probably on the wrong equipment.
The debate over “which strategy is better” is a false one. In its O&M Best Practices Guide, the U.S. Department of Energy defines Reliability-Centered Maintenance (RCM) precisely as “a strategic combination” of the approaches above: corrective, preventive, and predictive maintenance coexisting within the same plant, assigned asset by asset.
In other words, the right question isn't which strategy to adopt, but which strategy to assign to each asset—and based on what criteria. This article provides that decision matrix.
The Four Strategies, Without the Marketing Spin
1. Corrective (Run-to-Failure)
The DOE puts it plainly: “run it till it breaks.” No action is taken in advance; maintenance is performed only after the equipment has failed.
Pros: Minimal upfront cost and staffing.
Cons: Unplanned downtime, higher labor and spare-parts costs, and the risk of secondary damage to adjacent equipment.
When it is a legitimate decision—not negligence: for non-critical, redundant assets that are inexpensive to replace and have no safety or environmental consequences. Examples include a light fixture, an office fan, or a pump with an installed backup.
Choosing corrective maintenance is management. Relying on it by default is “firefighting mode.”
2. Time-Based or Usage-Based Preventive Maintenance
Tasks performed on a fixed schedule—every 3 months, every 500 hours, or every 10,000 cycles—regardless of the equipment's actual condition.
Pros: Easy to schedule and audit; suitable for consumables and components whose wear depends on usage.
Cons (as the DOE explicitly states): catastrophic failures still occur, it is labor-intensive, and it includes unnecessary maintenance—servicing equipment that is still in good condition.
Here's the point almost nobody mentions in a board meeting: the maintenance intervention itself introduces risk. Every disassembly and reassembly creates an opportunity to install a seal incorrectly, overtighten a bolt, or contaminate a clean bearing. It's self-inflicted infant mortality.
3. Condition-Based Maintenance (CBM)
A physical parameter of the equipment—vibration, temperature, ultrasound, oil analysis, or current—is measured and compared against a threshold. Maintenance is performed when the parameter crosses that threshold.
The applicable standard is ISO 17359:2018, Condition monitoring and diagnostics of machines: General guidelines, which defines how to set up the program: which parameters to measure, how to establish alarms, and how to diagnose problems.
CBM answers the question: What condition is this asset in today?
4. Predictive Maintenance (PdM)
Predictive maintenance is CBM plus trend analysis and modeling. It doesn't just detect that a parameter has deteriorated; it projects the curve and estimates how much time remains before functional failure.
PdM answers: How much time do I have left, and when should I intervene?
The distinction is clearer than the market suggests. In their Predictive Maintenance 4.0 study of 268 industrial companies in Belgium, Germany, and the Netherlands, PwC and Mainnovation define four maturity levels. The difference between levels 3 and 4 is precisely the boundary between CBM and predictive maintenance:
Level 3: Continuous real-time monitoring with alerts based on predefined rules or thresholds → this is CBM.
Level 4: Continuous monitoring with alerts based on predictive techniques, such as regression analysis → this is PdM.
One candid note about “prescriptive maintenance,” which you'll see in many catalogs: it is market terminology, not a standards-based term. There is no ISO or DOE definition behind it. Treat it as an emerging category, not a standard.
Want to know your plant's current maturity level? In a 45-minute assessment session, we'll review your asset base and provide a no-cost classification using PwC's maturity levels. Request an assessment.
The Data Point That Changes the Conversation with the CFO
The DOE publishes a comparison of annual maintenance costs per installed horsepower, based on the plant's dominant strategy:
Dominant strategy | Approximate cost |
Reactive | USD 18 / hp / year |
Preventive | USD 13 / hp / year |
Predictive | USD 9 / hp / year |
RCM (optimized mix) | USD 6 / hp / year |
The absolute values come from the guide's original edition and have not been updated to current prices. What matters is the ratio: a reactive plant spends approximately three times as much as a plant with mature RCM to maintain the same asset base.
The same document estimates that a preventive program saves 12% to 18% compared with a purely reactive one, and that a predictive program delivers an additional 8% to 12% in savings over preventive maintenance.
Deloitte's Making Maintenance Smarter provides the operational figures: predictive maintenance reduces maintenance planning time by 20% to 50%, increases availability by 10% to 20%, and reduces total maintenance costs by 5% to 10%. One often-overlooked point: Deloitte estimates that a poor maintenance strategy reduces a plant's production capacity by 5% to 20%. This isn't a maintenance-cost problem; it's a problem of installed capacity you've already paid for but aren't using.
Criterion 1: Criticality (Consequence × Probability)
The first filter is not technical; it's a business consideration. ISO 55000:2024 defines asset management as maximizing the value assets deliver to an organization's objectives throughout their life cycle. Put in shop-floor terms:
Criticality = consequence of failure × probability of occurrence.
The consequence should be broken down into five dimensions, each scored separately:
Personnel safety.
Environmental or regulatory impact.
Lost production (Is it a bottleneck? Is there redundancy?).
Quality (Would deterioration lead to scrap or rework?).
Repair cost and spare-parts availability.
Equipment that scores high on safety or environmental impact cannot be assigned a corrective-only strategy, regardless of monitoring cost. Equipment that scores low on all five dimensions does not need a sensor, no matter how much the vendor insists.
Criterion 2: The P-F Curve (The Test That Determines Whether CBM Is Viable)
This is the technical criterion that separates an informed decision from a purchase driven by trends.
The P-F curve describes the interval between point P—when a potential failure first becomes detectable—and point F—when deterioration progresses into functional failure.
The rule of thumb: inspection or measurement frequency must be less than half the P-F interval. If you measure monthly for a failure with a three-week P-F interval, you'll never detect it in time.
This leads to the article's most useful conclusion: if the P-F interval for a failure mode is shorter than your practical measurement capability, CBM won't work for that failure mode. You need redesign, redundancy, or instantaneous protection..
Typical P-F Intervals:
Failure mode | Typical P-F interval | Viable strategy |
Bearing wear | 2 to 10 weeks | CBM / predictive maintenance with weekly or continuous measurement |
Imbalance or misalignment | Weeks to months | CBM with periodic measurement |
Electrical insulation degradation | Months to years | Periodic inspection + thermography |
Electrical arc fault | Milliseconds to seconds | No inspection is sufficient: continuous protection or redesign |
The order in which technologies “see” a failure along this curve—from earliest to latest—is: ultrasound → vibration → oil analysis → thermography → electrical anomalies → performance decline (flow, pressure, production). By the time the operator notices that the pump “isn't performing like it used to,” you're in the final stretch of the curve, with the intervention window almost closed.
Do you know the P-F intervals for your three most critical assets? We can determine them with you in a 30 minutes technical workshop alongside your maintenance team. Schedule a workshop.
Criterion 3: Failure Patterns (Why Calendar-Based Maintenance Fails More Often Than You Think)
The hidden assumption behind all calendar-based maintenance is that machines wear out predictably as they age. For most failure modes, that assumption is false.
Classic studies of failure patterns—reviewed by Ron Moore and Douglas Plucknette on Reliabilityweb—conclude that the conditional probability of failure remains constant over time in five of the six recognized failure patterns. Depending on which study is used as the reference, this represents 83% to 97% of the equipment covered. In other words, random behavior, not age-related wear, is the dominant pattern.
The practical implication is uncomfortable: for most of your equipment, replacing a component “by the calendar” neither prevents failure nor makes use of its remaining useful life. It does help with the small percentage of components that do exhibit age-related wear—filters, belts, oil, and sacrificial elements—and in those cases, calendar-based preventive maintenance is the right, low-cost answer.
The Assignment Matrix
With all three criteria on the table, assigning a strategy is no longer a matter of opinion:
Asset type | Assigned strategy | Why |
Non-critical, redundant, low replacement cost | Planned corrective maintenance | The consequences are minor; monitoring costs more than the failure itself. |
Consumables and usage-wear components (filters, belts, oil, sacrificial elements) | Time-based or usage-based preventive maintenance | This is the segment where the failure pattern is age-dependent. |
Critical rotating equipment above 15 kW: motors, pumps, fans, compressors, gearboxes | CBM → predictive maintenance | P-F interval measured in weeks; vibration monitoring detects imbalance, misalignment, looseness, and bearing issues. |
Bottleneck asset in continuous production | Predictive maintenance + formal RCM | Every hour of downtime has a cost; a full analysis is justified. |
Instantaneous failure with no prior warning (electrical arc, sudden overpressure) | Redesign, redundancy, or protection—not inspection | The P-F interval is shorter than any measurement cycle. |
Obsolete end-of-life equipment with replacement already approved | Corrective maintenance + minimal monitoring | There is no point investing in an asset that is about to leave the asset base. |
The Mistake of Jumping Straight to Level 4
The PwC and Mainnovation study provides a picture of the industry's actual maturity: 25% of companies are at level 1 (visual inspection), 42% at level 2 (inspection with portable instruments), 20% at level 3 (continuous monitoring with thresholds), and just 11% at level 4 (model-based predictive maintenance).
That 11% didn't get there by skipping steps. Companies that did implement level 4 report concrete results: 95% achieved tangible outcomes, with an average +9% in uptime, −12% in costs, −14% in safety and environmental risk, and +20% in asset life.
McKinsey adds a warning that can prevent disappointment: low-maturity predictive maintenance programs deliver only a fraction of the potential benefit—around 10% or less. A sensor without a decision-making process behind it doesn't generate savings; it generates data.
Want a phased migration plan rather than a leap into the unknown? We'll build a 12-month roadmap with prioritized assets and a target maturity level for each. Request a roadmap.
What This Looks Like in a Colombian Plant
The local context reinforces the argument. According to ANDI's Joint Industrial Opinion Survey, Colombian manufacturing output fell 0.9% in 2024, and average installed-capacity utilization was 78.2%, down from 79.0% in 2023 and 80.5% in 2022.
With idle capacity and weak demand, every hour of unplanned downtime hurts more, not less: adding an extra shift to recover lost production is not easy, and there is less room to absorb maintenance cost overruns.
A realistic strategy-assignment plan for a medium-sized Colombian plant looks like this:
Weeks 1–2: Inventory assets and score their criticality across the five dimensions. This produces a shortlist: typically, 8 to 20 pieces of equipment account for most of the risk.
Weeks 3–4: For that shortlist, define the dominant failure modes and estimate the P-F intervals.
Month 2: Instrument assets where the P-F interval is measured in weeks and the consequences are high. Start at level 3 (thresholds), not level 4.
Months 3–6: Build a baseline. Without an asset-specific baseline, generic thresholds generate false positives and the team stops trusting the alerts.
Month 6 onward: Move to trend analysis and modeling for assets with enough historical data.
Conclusion
There is no single correct maintenance strategy for an entire plant. What matters is assigning the right strategy to each asset, based on three criteria: criticality (what happens if it fails?), the P-F curve (can I detect the failure in time?), and relative cost (does monitoring cost less than the failure?).
A plant that applies predictive maintenance to everything wastes money. One that applies time-based maintenance to everything wastes useful life and still experiences surprise stoppages. One that relies entirely on corrective maintenance lives in firefighting mode and pays three times as much to maintain the same asset base.
The difference between these three approaches isn't the budget. It's having done the asset-by-asset assignment exercise once, and done it thoughtfully.
Ready to assign a strategy to each asset in your plant? Schedule a 30-minute session, and we'll leave with a criticality matrix for your 10 most important assets. Schedule a session.
References and Sources
U.S. Department of Energy / PNNL — O&M Best Practices: Maintenance ApproachesU.S. Department of Energy / PNNL — O&M Best Practices: Maintenance Approaches
U.S. DOE / FEMP — O&M Best Practices Guide, Release 3.0, Cap. 5: Types of Maintenance Programs (PDF)U.S. DOE / FEMP — O&M Best Practices Guide, Release 3.0, Chapter 5: Types of Maintenance Programs (PDF)
Deloitte Insights — Making Maintenance Smarter: Predictive Maintenance and the Digital Supply Network (2017)Deloitte Insights — Making Maintenance Smarter: Predictive Maintenance and the Digital Supply Network (2017)
PwC & Mainnovation — Predictive Maintenance 4.0: Beyond the Hype (2018, PDF)PwC & Mainnovation — Predictive Maintenance 4.0: Beyond the Hype (2018, PDF)
McKinsey & Company — A smarter way to digitize maintenance and reliability (2021)McKinsey & Company — A Smarter Way to Digitize Maintenance and Reliability (2021)
ISO 17359:2018 — Condition monitoring and diagnostics of machines: General guidelinesISO 17359:2018 — Condition Monitoring and Diagnostics of Machines: General Guidelines
ISO 55000:2024 — Asset management: Vocabulary, overview and principlesISO 55000:2024 — Asset Management: Vocabulary, Overview and Principles
Reliabilityweb — RCM Failure Charts: Age Related or Random (Ron Moore & Douglas Plucknette)Reliabilityweb — RCM Failure Charts: Age Related or Random (Ron Moore & Douglas Plucknette)
ANDI — Informe Encuesta de Opinión Industrial Conjunta, diciembre 2024 (PDF)ANDI — Joint Industrial Opinion Survey Report, December 2024 (PDF)
Article prepared for Smartic — Industry 4.0, Industrial IoT, and Predictive Maintenance. Want to see Smartic in action at your plant? Schedule a 30-minute demo.




Comments