Preventive maintenance acts on a time or usage interval; predictive maintenance acts on evidence that indicates changing asset condition. Neither is automatically superior. The right policy depends on failure consequence, failure pattern, detectability, data quality, intervention cost, and the organization’s ability to act on a warning.
The short answer
Use scheduled preventive work where age, cycles, regulation, known wear, or a safe inspection interval gives a reliable basis. Consider predictive monitoring where condition changes can be measured early enough to support a useful decision and where avoided failure justifies sensors, integration, analysis, validation, and sustainment. Run-to-failure can still be rational for low-consequence, cheap, quickly replaceable assets.
- Choose policy per failure mode, not per asset name or technology trend.
- Quantify consequence, probability, warning interval, intervention lead time, and cost before adding monitoring.
- Validate sensor placement, context, timestamps, operating regimes, labels, and false-alarm behavior.
- Connect every detected condition to an owned maintenance decision and closed work order.
Decision guide
| Approach | Best fit | Evidence needed |
|---|---|---|
| Preventive | Age- or usage-related failure, mandated inspection, stable task and interval | Failure history, interval rationale, task quality, findings, cost, missed and unnecessary work |
| Predictive | Detectable condition change with useful warning and actionable intervention | Signal quality, operating context, baseline, threshold or model validation, lead time, decision outcomes |
| Run to failure | Low consequence, contained failure, economical replacement and available spare | Consequence review, spare and repair time, isolation plan, downstream impact |
| Redesign | Unacceptable consequence or recurring failure not controlled by maintenance | Root cause, risk reduction, design verification and management of change |
A practical workflow
- 01
Rank assets and individual failure modes by safety, environment, quality, delivery, and cost consequence.
- 02
Review failure history, duty, current tasks, parts, downtime, false diagnoses, and data availability.
- 03
Select a defensible policy and define the trigger, action owner, response window, and acceptance measure.
- 04
Pilot predictive methods on representative operating conditions and compare decisions against verified findings.
- 05
Review avoided loss, additional work, false alarms, missed failures, sustainment cost, and change the policy when evidence changes.
Evidence and acceptance
For each covered failure mode, preserve the selection rationale, baseline, data lineage, operating context, task or trigger, competence requirement, work orders, inspection findings, parts, downtime, costs, exceptions, and periodic review. Predictive value is established by better maintenance decisions—not merely by collecting more signals or plotting a trend.
Common mistakes
- Calling a fixed alarm threshold predictive maintenance without validating it against condition and operating state.
- Installing sensors before defining the failure mode and maintenance decision.
- Training or tuning on unlabelled events, shifted timestamps, or one narrow operating regime.
- Ignoring the labor, calibration, connectivity, cybersecurity, and model-maintenance cost.
- Producing alerts that no named person can convert into timely work.
Engineering and security boundary
Primary sources
Technical claims and product boundaries were checked against these primary sources. Fact-checked
Frequently asked questions
Is predictive maintenance always cheaper?
No. Sensors, engineering, connectivity, analysis, validation, support, and false alarms can cost more than a well-designed preventive policy for some failure modes.
How much history is needed?
There is no universal duration. It depends on failure frequency, operating regimes, sensor quality, labels, seasonality, and the decision being validated. Sparse failures require especially cautious claims.
Can one asset use several maintenance policies?
Yes. Different failure modes on the same asset may justify scheduled work, condition monitoring, functional testing, run-to-failure, or redesign.
Does AutoNexit predict equipment failures?
No. It organizes the project and maintenance evidence around decisions; monitoring and diagnostic systems perform runtime analysis.