Preventive vs Predictive Maintenance: Where CMMS Fits

Pump P-104 has a scheduled lubrication task. It receives periodic vibration readings. Operators sometimes report a change in sound or temperature. A condition alert may trigger investigation. A bearing may be replaced before the pump can no longer perform its function. Another protective function may still require periodic testing.

Is P-104 preventive or predictive?

Often, “versus” is the wrong mental model.

An asset does not necessarily have one maintenance strategy. Different ways it can fail may justify different ways of deciding when to intervene.

Preventive and predictive maintenance are different evidence strategies for deciding when intervention is justified. Preventive relies more heavily on predetermined trigger logic. Predictive relies more heavily on current condition evidence. Neither is inherently superior. Neither eliminates uncertainty. Both should learn from what maintenance actually finds.

The CMMS fits at the boundary where maintenance evidence becomes accountable work, and where the physical findings from that work become evidence for whether the strategy was right.

Maintenance strategies are not rungs on a ladder

A common slide shows maintenance as a maturity ladder: reactive, then preventive, then predictive, then prescriptive. That picture can be useful as a warning against unmanaged breakdown maintenance, but it becomes misleading when buyers treat it as a universal progression.

NASA’s current facilities maintenance guidance describes reliability-centered maintenance as an approach that seeks an optimal mix of proactive and reactive maintenance, gathers performance data, and integrates strategies according to their strengths and constraints.[1]

That point matters for software selection. A maintenance program can be sophisticated and still use recurring PM tasks, condition monitoring, operator observations, inspections, and corrective work on the same asset. The question is not which label sounds most advanced. The question is which evidence is appropriate for each failure risk.

Maintenance strategies are different ways of managing different failure risks.

Start with the maintenance trigger

The shallow distinction says preventive equals scheduled work and predictive equals unscheduled work. That misses the real difference.

Predictive maintenance may include scheduled observation. A vibration route may happen every month. Oil samples may be collected on a defined cadence. Thermography may be performed during a planned inspection window. The intervention, though, is justified by interpreted condition evidence rather than by the calendar alone.

Preventive maintenance may also include inspection rather than replacement. A scheduled inspection can create evidence that leads to work, or it can create evidence that no action is needed.

The better buyer question is: what evidence caused this intervention to happen now?

CMMSBuyersGuide model: the Maintenance Trigger Model

This is a CMMSBuyersGuide explanatory model, not an industry standard. Its purpose is to separate the evidence that starts maintenance from the work that follows.

PREDETERMINED TRIGGER LOGIC Calendar / runtime / cycles / usage / justified recurrence ↓ Maintenance or inspection activity ↓ The organization decided beforehand that reaching this trigger justifies the activity.

CONDITION EVIDENCE Observe or measure ↓ Condition indicator ↓ Interpret ↓ Decide whether action is warranted ↓ Maintenance activity when justified.

FAILURE OR EVENT TRIGGER Functional failure or other event occurs ↓ Corrective response.

This model is intentionally narrow. It is not a complete maintenance taxonomy. It asks one disciplined question: what evidence caused this work, inspection or response to happen now?

Related guide: Preventive Maintenance Software

Preventive and predictive place uncertainty in different places

Preventive maintenance does not eliminate uncertainty. A central uncertainty is whether the interval, runtime trigger or other predefined recurrence is appropriate for the failure mode and operating context.

If the trigger is too frequent, maintenance may be performed with limited benefit. If the trigger is too infrequent, meaningful deterioration may occur between interventions. A recurring task can look orderly while still resting on an unexamined assumption.

Predictive maintenance does not eliminate uncertainty either. It shifts more of the decision toward current condition evidence. The organization must decide whether a measurement, observation, alert or trend means action is warranted.

Preventive maintenance asks whether the trigger interval is right. Predictive maintenance asks whether the condition evidence is right.

That is an explanatory contrast, not a legal definition. Many real programs combine interval-based, inspection-based and condition-based logic.

Predictive maintenance can move the schedule from intervention to observation

P-104 might receive vibration measurement every month. The monthly route is scheduled. The bearing is not automatically replaced every month.

EVERY MONTH ↓ Observe condition ↓ Interpret ↓ Intervene only if warranted.

Predictive maintenance can move the schedule from intervention to observation. Its purpose can be to schedule maintenance at a better-informed time.

NASA’s predictive testing and inspection appendix describes techniques such as vibration analysis, lubricant analysis, thermography and other inspection methods as ways to observe equipment condition.[2]

For buyers, this creates a different workflow requirement. The software must respect the distinction between the monitoring task and the maintenance action that may follow.

Related guide: Maintenance Scheduling Software

Observation frequency is itself a maintenance decision

If condition monitoring is periodic rather than continuous, someone still has to decide how often to look. Monthly observation, weekly observation and continuous monitoring create different burdens and different chances of seeing deterioration in useful time.

Monitoring too rarely can miss a useful deterioration window. Monitoring far more often than needed can add cost, data volume and review work without materially improving the decision.

FAILURE BEHAVIOR ↔ DETECTABILITY ↔ OBSERVATION FREQUENCY ↔ RESPONSE TIME

Predictive maintenance can replace a fixed intervention interval with an observation interval, and the observation interval still has to make sense.

More frequent monitoring is not automatically better monitoring. Buyers should ask how the monitoring cadence relates to the failure behavior, how quickly the organization can interpret a signal, and how long it needs to prepare a useful response.

Use the P-F concept carefully

The P-F idea is useful when it is treated as a practical warning concept rather than a precise promise. In plain terms, deterioration may become detectable before the asset reaches a functional failure or an unacceptable consequence.

Degradation becomes detectable ↓ Potential failure or detectable condition ↓ Time passes ↓ Functional failure or unacceptable consequence

The buyer question is not only how early a system can detect something. It is how much of that warning interval the maintenance organization can actually use.

CMMSBuyersGuide model: the Usable Warning Window

This is a CMMSBuyersGuide explanatory model, not a mathematical reliability formula or standard metric.

DETECTION ↓ INTERPRET ↓ DECIDE ↓ PREPARE ↓ GAIN ACCESS / SCHEDULE ↓ INTERVENE ↓ Before unacceptable failure consequence

Usable Warning Window means the portion of an early warning that the maintenance organization can actually use to understand, prepare and intervene.

A warning is only early if enough useful time remains to act on it.

Suppose a condition becomes detectable roughly 30 days before likely functional failure. That does not mean the maintenance team has 30 useful days. The alert may wait for review. Diagnosis may take time. A part may need to be ordered. Production access may be constrained. Work may need a shutdown window.

Detection lead time and usable maintenance time are different things.

Some of the warning window can disappear before the organization even knows the warning exists. That is why predictive maintenance is partly a planning and accountability problem, not only a sensing problem.

Related guide: Maintenance Planning Software

Condition data is evidence, not the machine itself

A vibration sensor measures vibration. A temperature sensor measures temperature. Oil analysis measures properties of a sample. Inspection creates observations. Those observations must be interpreted in relation to equipment condition and operating context.

NIST’s review of diagnostics and prognostics describes PHM systems as using real-time and historical state information to provide actionable information for decision-making, while also noting that the field has implementation challenges and needs around verification, validation, data management and economics.[3]

Condition data is evidence about the machine. It is not the machine itself.

PHYSICAL CONDITION ↓ Measurement or observation ↓ Condition indicator ↓ Interpretation ↓ Maintenance decision

This is why a buyer should be wary of any workflow that treats a signal as a complete maintenance answer. A signal can be valuable, but it still needs interpretation, ownership and a decision path.

Detection and action are different decisions

A condition indicator changes. That does not automatically mean replace the component now.

Detection means something changed enough to become observable. Action means the evidence is strong and relevant enough that maintenance intervention is justified. Between those two points, the organization may need review, comparison with history, inspection, operating-context checks or reliability judgment.

Detection creates information. Maintenance requires a decision.

Some workflows can automate parts of this path when the evidence is well understood and the consequence of delay is unacceptable. Other workflows need human interpretation. The important point is clarity: where is the maintenance decision being made, and can the buyer understand why?

Monitoring is imperfect

Condition monitoring can be useful and still be imperfect. The practical issue is not whether every signal can be labeled perfectly. It is whether the program creates enough confidence to support better maintenance decisions.

Peer-reviewed work on imperfect inspections in condition-based and predictive maintenance explicitly models correct and incorrect inspection decisions, including true-positive, false-positive, true-negative and false-negative outcomes.[4]

A false or low-value positive can send people to investigate a condition that does not confirm the assumed maintenance need. A missed condition can create false confidence or a lost opportunity to intervene before consequence. Many real cases are messier than a clean true/false label, but the tradeoff is real.

A predictive program must detect change and create enough confidence in what the change means to support a maintenance decision.

Thresholds create tradeoffs

Thresholds and alert rules are decision devices. If a threshold is very sensitive, it may create many alerts. If it is too broad, meaningful conditions may be missed. A threshold also determines how much uncertainty the maintenance organization must investigate.

Research on imperfect inspection intervals shows that inspection error probabilities can vary with operating time, and that assuming constant errors can distort maintenance-policy results in the studied model.[5]

Sensitivity and usefulness are not the same thing.

This guide does not prescribe vibration, temperature, oil-analysis or inspection thresholds. The buyer issue is whether the system can carry the decision context: why an alert mattered, who reviewed it, what was decided, what work followed, and what the technician actually found.

More alerts can create maintenance information work

Condition monitoring can reduce some unnecessary physical interventions while increasing the amount of information that must be interpreted. That creates a possible condition review backlog: condition evidence awaiting maintenance disposition.

This is a CMMSBuyersGuide explanatory concept. It does not mean every organization formally maintains such a queue. It means predictive maintenance can create information work that has to be owned.

• Is the signal real?
• Is it meaningful?
• Should we monitor, inspect, act, dismiss or escalate?
• Who owns the next decision?
• What happens if the same condition appears again?

Predictive maintenance can reduce physical maintenance while increasing the amount of maintenance information that requires interpretation.

Signal ownership is a control issue

A monitoring platform may create an alert. The next decision may belong to reliability, a planner, a supervisor, a technician or an external monitoring provider. The right owner depends on the organization and the risk.

The buyer question is practical: who is accountable for deciding what happens next?

A detected condition without ownership is not yet controlled maintenance.

If nobody owns disposition, predictive maintenance can become a second inbox beside the CMMS. If every alert becomes a work order automatically, the CMMS can fill with work that has not yet been justified. The useful middle ground is a workflow where uncertainty can be reviewed without disappearing.

Predictive maintenance needs somewhere for uncertainty to live

A signal appears. The organization may not know yet whether maintenance work is justified. It may need a review state, an investigation state, a decision to continue monitoring, or an action-required decision. The exact status names can vary.

CONDITION OBSERVED ↓ UNDER REVIEW ↓ Continue monitoring / investigate / action required / no action with reason ↓ Work, if justified

Predictive maintenance needs somewhere for uncertainty to live between “something changed” and “we have decided to intervene.”

Two extremes are risky. One turns every alert into a work order before the maintenance decision is clear. The other leaves alerts in a separate dashboard where they never become controlled maintenance.

Related guide: Predictive Maintenance Software

Automatic work order generation is not automatically sophisticated

Automatic work order generation can be useful. It can also hide a weak decision process. The question is where the maintenance decision is being made and whether the organization can understand why.

SIGNAL ≠ CONFIRMED CONDITION ≠ MAINTENANCE NEED ≠ APPROVED WORK

Some signals may be mature enough to generate work directly. Others may require review, inspection or approval. Buyers should test whether the system can support both cases without losing the evidence trail.

CMMSBuyersGuide model: the Prediction-to-Work Chain

This is a CMMSBuyersGuide synthesis. It shows how condition evidence becomes accountable maintenance.

OBSERVE ↓ DETECT ↓ INTERPRET ↓ DECIDE ↓ OWN ↓ PLAN ↓ PREPARE ↓ SCHEDULE / FIND OPPORTUNITY ↓ EXECUTE ↓ VERIFY ↓ FEEDBACK

Predictive maintenance does not end when software predicts something. That is where maintenance begins.

Condition monitoring creates evidence. Maintenance management creates accountability for what the organization does with that evidence.

Where the CMMS fits

The CMMS does not have to perform the prediction to be essential to predictive maintenance. It also should not be assumed to replace the sensor platform, historian, SCADA system, OEM diagnostic tool, vibration-analysis package, oil-analysis lab, reliability-engineering tool, ERP or production scheduling system.

Condition and reliability systems may own observation, detection and analysis. The CMMS may become the place where evidence becomes an owned maintenance need, planned work, scheduled work, executed work, physical findings, maintenance history and feedback.

Condition / reliability systems Observation → detection → analysis ↓ CMMS boundary Owned maintenance need → planned work → scheduled work → executed work → physical finding → maintenance history → feedback

The CMMS can become the bridge between why maintenance was triggered and what maintenance actually discovered.

Related guide: Maintenance Tracking Software

Predictive value is created by detection plus response

A technically accurate warning alone does not guarantee maintenance value. An algorithm may correctly identify bearing degradation, while the alert is not reviewed, no owner is assigned, the part is unavailable, production access is unavailable, and work is never prepared. In that case, the prediction may have succeeded analytically while the maintenance loop failed operationally.

Prediction accuracy and maintenance usefulness are related, but they are not identical.

Predictive maintenance value is created by a system of detection and response, not by detection alone.

Economics: be skeptical of universal ROI percentages

Buyers should be skeptical of universal predictive-maintenance ROI percentages. Results depend on asset population, failure behavior, consequence, current practice, monitoring technology, implementation quality, labor, operating environment, response capability and other local factors.

NIST’s machinery-maintenance economics work discusses costs, losses and benefits in manufacturing while emphasizing variation in estimates, data gaps and the need to evaluate maintenance investments in context.[6]

A recent NIST systematic review of condition-monitoring-based technologies found that evaluation studies vary in how they characterize applications, monitoring, maintenance deployment, performance measures and economic analysis.[7]

The cost of knowing is an explanatory phrase, not formal accounting terminology. Condition monitoring can involve sensors, installation, connectivity, data infrastructure, integration, analysis, specialist time, alert review, investigation, system upkeep, training and support. Potential value may include better timing, fewer unnecessary interventions, improved planning, better parts readiness and better coordination.

Condition monitoring buys information. The economic question is whether that information improves maintenance decisions enough to justify what it costs to obtain and act on it.

The value of condition monitoring is the decisions that become better because the data exists.

Monitor everything is not a strategy

The ability to collect condition data is not itself a reason to collect condition data. The evaluation should happen at the relevant failure mode, consequence and operating context.

A low-consequence failure with no useful detectable precursor may not justify sophisticated monitoring. A failure mode with clear condition evidence, enough warning time, practical response options and meaningful consequence may deserve a different treatment.

CMMSBuyersGuide framework: Predictive Suitability Test

This is a CMMSBuyersGuide framework for buyer discussions. It is not a numeric score.

1. Observable? Does degradation create useful evidence?
2. Detectable early enough? Can the condition become visible with useful lead time?
3. Practical to observe? Can it be monitored reliably at an appropriate cadence?
4. Interpretable? Can the evidence support a meaningful maintenance decision with acceptable uncertainty?
5. Actionable? Can the organization actually do something useful after detection?
6. Consequential? Does earlier intervention materially change an important outcome?
7. Economic? Does the improved decision justify the monitoring burden and cost?

Where does better condition information materially improve the maintenance decision?

Same asset, different failure modes, different strategies

Return to P-104. A lubrication need might follow justified recurring logic. Bearing degradation might be monitored through vibration. Seal leakage might be discovered through inspection or operator observation. A hidden or protective function might require periodic testing. A failure with no useful detectable precursor may gain little from predictive monitoring.

This is an illustrative example, not an engineering prescription for that fictional pump.

“Is this asset preventive or predictive?” may be the wrong question. Ask which failure mode you are trying to manage and what evidence can tell you when intervention is justified.

Preventive, predictive and corrective can describe different layers of the same maintenance program rather than competing alternatives. The same bearing could involve scheduled lubrication, scheduled condition monitoring and condition-triggered corrective replacement.

Related guide: Asset Maintenance Management

Corrective does not necessarily mean emergency

NIST-linked maintenance-terminology research argues that PHM can initiate maintenance after a failure process has begun and before functional failure, creating corrective work that familiar metrics may misread.[8]

That means one job can be predictive or condition-triggered in strategy, corrective in purpose, and planned in execution. A condition alert may detect an incipient bearing problem. The resulting work may correct that condition. The work may still be planned, scheduled and completed before functional failure.

Maintenance strategy, maintenance action and work execution are different dimensions.

One work order can carry three different truths. Reporting that collapses them into a single label can mislead buyers and managers.

Predictive can make old KPIs misleading

A site may try to reduce its corrective-maintenance percentage. Then a predictive program finds more incipient defects and creates corrective work before functional failure. Corrective work rises. That does not automatically mean performance got worse. It may mean the organization is finding problems earlier.

The reverse can also be true. A rising number of condition-triggered corrective jobs could reflect noisy alerts, weak thresholds or poor strategy. The label alone is insufficient.

Predictive maintenance can change the meaning of familiar maintenance metrics.

A work-order label can describe what kind of work was performed without proving whether the maintenance strategy succeeded.

Trigger provenance: preserve why the work exists

Consider two work orders with the same instruction: replace bearing. Work Order A exists because a 500-hour interval was reached. Work Order B exists because condition evidence indicated deterioration.

The physical intervention may be identical. The maintenance logic is different. If the CMMS stores only that the bearing was replaced, it loses part of the strategy record.

A maintenance record that preserves what was done but not why it became necessary loses part of the maintenance strategy.

Implementation can vary. Trigger provenance might be preserved through a linked alert, PM schedule, inspection finding, operator observation, condition event, related request, parent-child work relationship or another traceable mechanism. The principle matters more than the field name.

What was done is only half the story. Why the organization decided to do it is part of the maintenance record too.

Condition decisions deserve history

Predictive maintenance creates decisions before it creates work. A signal may be reviewed and dispositioned as continue monitoring, investigate, act, no action with reason, or another appropriate outcome. These are examples, not prescribed software statuses.

Condition evidence should not disappear merely because it did not become a work order. If the same signal appears tomorrow, the reviewer should ideally know it was already considered and what was decided.

Predictive maintenance needs a memory of the work it created and the condition evidence it deliberately chose not to act on.

This creates a bridge between condition history and work history. It also gives reliability and maintenance leaders a way to review whether signals are repeatedly low-value, useful, late, ignored or poorly connected to findings.

Preventive maintenance has hidden decisions too

A recurring PM every 500 hours may execute a maintenance decision originally made years earlier. The source might be OEM guidance, engineering analysis, experience, regulatory requirement, historical practice or an unknown legacy decision.

Old intervals are not automatically wrong. The problem is losing the rationale. When a recurring trigger continues for years, buyers need enough context to know whether the decision still matches the asset, failure mode and operating environment.

A recurring PM interval is a maintenance decision repeated automatically.

Automation does not eliminate the need to understand the decision being automated.

Predictive strategy can become stale too

Predictive is not automatically dynamic or adaptive. A condition threshold could have been configured years ago. Since then, asset duty may have changed, operating context may have changed, the sensor may have changed, the baseline may have changed, or model assumptions may have changed.

A condition threshold can become just as unquestioned as a calendar interval.

A strategy is not adaptive merely because its input is digital.

Digital does not mean adaptive. Buyers should ask how threshold rationale, model assumptions, review decisions and maintenance findings are preserved across time.

CMMSBuyersGuide model: the Strategy Learning Loop

This CMMSBuyersGuide synthesis applies to preventive and predictive maintenance.

TRIGGER LOGIC ↓ MAINTENANCE DECISION ↓ WORK ↓ PHYSICAL FINDING ↓ INTERVENTION ↓ ASSET OUTCOME ↓ REVIEW WHAT THE TRIGGER ACTUALLY TOLD US ↓ REFINE OR RETAIN STRATEGY AS JUSTIFIED

Preventive maintenance is a hypothesis: intervening or inspecting at this interval is justified. Repeated maintenance findings create evidence about that hypothesis. A quarterly inspection that repeatedly finds nothing actionable does not automatically prove the interval is wrong. It does create information for review.

Predictive maintenance is also a hypothesis: this condition evidence tells us something useful about when intervention is justified. After work occurs, the maintenance finding matters. Did the suspected condition exist? Was warning early enough? Was the intervention appropriate? Did the signal represent something else?

The maintenance finding is critical feedback about what the condition signal actually represented.

Technician findings can contain uncertainty too, so they should not be treated as perfect truth. They are still essential evidence. Predictive maintenance without maintenance feedback is an open loop.

What a CMMS should help preserve

This is not a feature checklist. Exact implementation can vary by system, integration and operating model. Conceptually, a useful maintenance record preserves relationships among the following:

ASSET / FAILURE CONTEXT ↓ MAINTENANCE STRATEGY ↓ TRIGGER / CONDITION EVIDENCE ↓ DECISION / DISPOSITION ↓ WORK ↓ PHYSICAL FINDING ↓ INTERVENTION ↓ OUTCOME ↓ STRATEGY LEARNING

The important issue is continuity. Can the organization follow the evidence forward into work, and follow completed work backward to the reason it existed?

Related guide: Plant Maintenance Software

Distinction wall

Each distinction has a workflow consequence. Buyers should test whether reporting, work history and integrations preserve these differences or flatten them into a simple preventive-versus-predictive label.

Buyer questions

• Can we distinguish why maintenance work was triggered?
• Can interval-generated work be distinguished from condition-triggered work?
• Can a condition signal exist without automatically becoming a work order?
• Where does condition evidence live while it is being reviewed?
• Can someone own the disposition of a condition signal?
• Can we record a decision to continue monitoring without losing the original evidence?
• Can a later signal be connected to an earlier review or decision?
• Can condition evidence be linked to the work it eventually creates?
• Can completed work be traced back to the evidence that triggered it?
• Can technician findings remain connected to the original alert?
• Can we distinguish scheduled monitoring from scheduled intervention?
• Can the system support multiple maintenance strategies on the same asset for different failure modes?
• Can scheduled lubrication coexist with condition monitoring and condition-triggered corrective work?
• Can we preserve the rationale or source behind a recurring PM trigger?
• Can we review what repeated preventive inspections actually found?
• Can we review what predictive alerts actually led technicians to find?
• Can we distinguish condition-triggered corrective work from emergency corrective work?
• Can reporting distinguish maintenance strategy, maintenance action and execution state?
• Can we see whether warning arrived with enough time to prepare useful work?
• Can planning, material and access constraints be connected to condition-triggered work?
• Can alerts that did not become work still have disposition history?
• Can we tell who is responsible for reviewing condition evidence?
• Can monitoring data come from external systems without forcing CMMS to replace those systems?
• Can the CMMS exchange useful status and outcome information with condition-monitoring systems where appropriate?
• Can we tell when a predictive trigger repeatedly produces low-value or unconfirmed findings?
• Can we tell when preventive work repeatedly produces no actionable findings?
• Can maintenance outcomes support later strategy review?
• Can trigger logic be updated without destroying historical context?
• Can we preserve why a threshold or interval exists?
• Can we distinguish current strategy from historical strategy?
• Does the system help us understand maintenance decisions, or merely count work orders?

Vendor demo tests

1. Same work, different trigger

Create two bearing-replacement work orders. One exists because a scheduled interval was reached. The other exists because a condition alert was reviewed. Ask the vendor to show how the system preserves why each work order exists. Six months later, can you tell which strategy generated each intervention and what maintenance found?

2. Alert that does not yet deserve work

Start with vibration increase on P-104. Ask what happens before this becomes a work order. Can the organization preserve the signal, owner, review, decision, continue-monitoring disposition, investigation and eventual action without forcing every alert into work or leaving it in a forgotten dashboard?

3. Trace the trigger

Start with a condition alert and follow it through signal, review, decision, work, finding, completion and outcome. Then reverse direction from the completed work order. Can the system show what evidence originally caused the work to exist?

4. Usable warning window

Create a bearing-condition warning that requires review, response decision, part availability, work preparation, access coordination and execution before unacceptable deterioration. Ask how the system helps turn the warning into prepared, owned work before useful response time disappears. The CMMS does not have to predict remaining useful life for this demo.

5. Low-value or unconfirmed alert

A condition alert indicates a suspected bearing issue. A technician investigates and the finding does not support the assumed failure mode. Ask where that finding goes, whether it remains connected to the original signal, and whether the reliability or condition process can learn from it.

6. Preventive task repeatedly finds nothing

Use a quarterly inspection with multiple executions and no actionable findings. Ask whether history is visible enough to evaluate whether the task and interval remain justified. The system should not automatically cancel or change the PM; the demo tests whether review evidence exists.

7. Condition signal with no work

A condition alert is reviewed and the decision is continue monitoring. Ask where that decision lives. Then trigger the same or similar condition later. Can the reviewer see that the condition was previously evaluated and what was decided?

8. KPI trap

Condition monitoring detects degradation and creates corrective bearing work that is planned and scheduled before functional failure. Ask how reporting represents it. Can the system distinguish why the work was triggered, what type of intervention occurred and whether execution was planned or emergency?

9. Predictive suitability

Pick several failure modes on P-104. Ask the vendor to show how strategy can differ by failure mode rather than forcing one strategy onto the entire asset. The system does not necessarily need a formal failure-mode strategy engine. The test is whether the relationship among asset, failure behavior, trigger, task or monitoring, work and finding can be preserved.

Related guide: Equipment Maintenance Software

Related guide: Mobile CMMS

Conclusion: ask what the intervention teaches

Do not ask whether P-104 should be preventive or predictive. Ask, for each important failure mode, what evidence can tell the organization when intervention is justified.

Maybe lubrication follows justified recurring logic. Bearing condition is monitored. Another condition is discovered by inspection. Another failure gives no useful predictive warning. Condition monitoring may produce corrective work that is still planned. Preventive inspection may itself generate condition evidence. The categories coexist.

The point is to use evidence appropriate to the failure risk and make sure the organization can act on what that evidence tells it.

The CMMS does not need to be the sensor, diagnostic model or condition-monitoring platform. Its critical role is preserving the path from maintenance trigger to decision to work to physical finding, and making that finding available to improve the next decision.

Preventive and predictive maintenance become much more useful when the question changes from “Which one is better?” to “What evidence should cause us to act, can we act in time, and what did the intervention teach us?”

Final buyer question: can your maintenance system tell you what work was performed, why it happened when it did, and whether what you found justified the strategy that triggered it?