Your facility runs bearings to failure. Not by policy, but by default: nobody watches the bearing until it breaks, and when it breaks, the line stops and the emergency starts. Everyone agrees predictive maintenance would be better. But the last time someone bought vibration equipment, it sat in a drawer after two months because nobody knew how to build a program around it.
That is the real problem. Predictive maintenance is not a tool you buy; it is a program you run. The instrument is maybe 20% of it. The other 80% is the process: what you monitor, how often, who reads the data, what triggers action, and how the finding becomes a scheduled repair instead of another emergency. The cost of skipping this is well documented in what run-to-failure actually costs once downtime and secondary damage are counted.
Here are the five steps to build a predictive maintenance program that actually works, instead of buying an instrument that ends up in a drawer.
Step 1: Rank Equipment by Criticality Before You Monitor Anything
The first mistake most programs make is trying to monitor everything, which spreads effort so thin that nothing is monitored well. The program collapses under its own weight, and the equipment gets abandoned.
The correct starting point is criticality ranking. Not every bearing deserves the same attention. A bearing whose failure stops the entire production line, creates a safety hazard, or costs $50,000 in downtime deserves intensive monitoring. A bearing whose failure is a minor inconvenience deserves little or none. Ranking equipment by consequence-of-failure focuses your limited monitoring effort where it returns the most.
How to do it: Score each equipment position by consequence of failure: production impact, safety, repair cost, and lead time to repair. Sort into tiers, critical, important, and general. This ranking is the foundation of the entire program; it determines where you invest monitoring effort and where you accept run-to-failure as a rational choice.
Step 2: Match the Monitoring Method to Each Equipment Tier
With equipment ranked, match the monitoring method to each tier. This is where most programs waste money, buying one technology and applying it everywhere, when different equipment needs different methods. The right method depends on the equipment’s speed, criticality, and failure modes, and choosing correctly is a subject in itself, covered in the detail on which monitoring method fits which application.
Critical high-speed equipment justifies permanent online vibration monitoring with automatic alerts. Important equipment justifies periodic portable vibration or ultrasonic surveys. General equipment justifies simple temperature checks and inspection rounds. Low-speed equipment needs ultrasonic or specialized methods because standard vibration analysis struggles below a few hundred RPM.
How to do it: Assign a monitoring method to each tier based on criticality and equipment characteristics. Permanent online monitoring for the critical few, periodic surveys for the important many, and basic checks for the general majority. Match the investment to the consequence, and match the technology to the equipment.
Step 3: Establish Baselines and Alarm Thresholds
A monitoring reading means nothing without a baseline. A vibration level of 4 mm/s is not inherently good or bad; it is good if the baseline is 4 and bad if the baseline is 1.5 and climbing. Programs that collect readings without baselines produce numbers nobody can interpret, and the data gets ignored.
Establishing baselines means measuring each monitored point when the equipment is known to be healthy, and recording that as the reference. Alarm thresholds are then set relative to the baseline: an alert level where the trend warrants attention, and an alarm level where action is required. Without these, the program produces data but no decisions.
How to do it: Measure and record a baseline for every monitored point on healthy equipment. Set alert and alarm thresholds relative to the baseline, using manufacturer guidance and standards as starting points. Refine the thresholds as you learn each machine’s normal behavior. The baseline turns a raw number into a meaningful signal.
Step 4: Assign Ownership and a Review Rhythm
This is the step that separates working programs from abandoned ones, and it has nothing to do with technology. Data that nobody owns and nobody reviews on a schedule is data that gets collected and forgotten. The instrument in the drawer failed here: somebody collected readings for a while, nobody was responsible for acting on them, and the effort faded.
A working program has a named owner for the data, a fixed review rhythm, and a clear escalation path. Someone reviews the readings on schedule, identifies the points that are trending toward alarm, and initiates action. The review is a calendar commitment, not a when-we-get-to-it task.
How to do it: Name a person responsible for reviewing predictive data. Set a fixed review cadence (weekly for critical online data, monthly for survey routes). Define what happens when a threshold is crossed: who is notified, who decides, and how the finding enters the work-order system. Ownership and rhythm are what keep the program alive after the initial enthusiasm fades.
Step 5: Convert Findings Into Scheduled Work
The entire point of predictive maintenance is to convert an unplanned failure into a planned repair. That conversion only happens if the monitoring finding flows into the maintenance planning system and becomes a scheduled work order with parts staged and time allocated.
Programs fail at this last step when a finding is noted but not acted on, or acted on as another emergency. The bearing is flagged as trending toward failure, and then nothing is scheduled, so it fails anyway, in production, exactly as it would have without monitoring. The monitoring worked; the response did not.
How to do it: When monitoring flags a developing problem, generate a work order with appropriate priority based on how much lead time the finding provides. Stage the parts, schedule the repair in a planned window, and close the loop by confirming the repair addressed the flagged condition. This is where predictive maintenance delivers its return: the failure that would have stopped the line becomes a scheduled two-hour repair during a planned shutdown.
Build a Program That Lasts, Not an Instrument That Sits
At IVOR Bearings, we help facilities build predictive maintenance programs, from criticality ranking to method selection to the review rhythm that keeps the program alive. Contact us at quotes@ivorbearings.com, call +1 305-769-9459, or visit our contact page HERE to discuss your predictive maintenance program.
Why Programs Fail: The Five Failure Points
Predictive maintenance programs fail at predictable points, each corresponding to a skipped step:
Monitoring everything, so nothing well: No criticality ranking. Effort spread too thin. The program collapses.
One technology for all equipment: No method matching. The wrong tool on the wrong equipment produces useless data.
Readings with no baseline: No reference. Numbers nobody can interpret. Data ignored.
Data nobody owns: No ownership or review rhythm. Readings collected and forgotten. The instrument goes in the drawer.
Findings that do not become work: No conversion to scheduled repair. The bearing fails in production despite being flagged.
A program that completes all five steps avoids all five failure points. Most abandoned programs skipped steps 1, 4, or 5, the process steps, while focusing only on step 2, the instrument.
The Financial Impact of a Working Program
A facility with 300 bearing positions, 40 of them critical, moving from run-to-failure to a working predictive program:
With run-to-failure:
- Unplanned bearing failures per year: 30-50
- Average cost per unplanned failure including downtime and secondary damage: $8,000-$25,000
- Annual cost: $240,000-$1,250,000
With a working predictive program:
- Failures caught and converted to planned repairs: 70-85%
- Unplanned failures remaining: 6-12 per year
- Program cost (instruments, software, labor): $30,000-$70,000/year
- Annual cost including program: $80,000-$370,000
Net annual savings: $160,000-$880,000, driven almost entirely by converting unplanned failures into planned repairs. The savings come from the process, not the instrument; the same instrument in a program without steps 1, 4, and 5 delivers almost none of this.
Your Action Plan
- Rank your equipment by criticality – Score every position by consequence of failure and sort into critical, important, and general tiers. This is step one and the foundation of everything.
- Match monitoring methods to tiers – Permanent online for critical, periodic surveys for important, basic checks for general. Match the technology to the equipment.
- Establish baselines before expecting to find problems – Measure healthy equipment and set thresholds relative to those baselines.
- Name an owner and set a review rhythm – This is the step that keeps the program alive. Data with an owner and a schedule survives; data without them does not.
- Partner with predictive maintenance experts – At IVOR Bearings, we help build the full program: criticality ranking, method selection, baselines, review rhythm, and the conversion of findings into scheduled work.
We provide: criticality ranking support, monitoring method selection matched to your equipment, baseline and threshold guidance, program structure and review rhythm design, and technical support interpreting findings and planning repairs.
Our customers convert the majority of their bearing failures from unplanned emergencies into scheduled repairs, and build programs that outlast the initial enthusiasm because the process, not just the instrument, is in place.
The Instrument Is Not the Program
Every abandoned predictive maintenance effort started with an instrument and skipped the process. The vibration analyzer in the drawer worked perfectly; what was missing was the criticality ranking that would have focused it, the baselines that would have made its readings meaningful, the owner who would have reviewed them, and the planning system that would have turned findings into repairs.
Build the program, all five steps, and the instrument becomes what it was meant to be: the sensor at the front of a process that converts failures into scheduled work. Buy only the instrument, and it becomes what it usually becomes: an expensive object in a drawer.
Contact IVOR Bearings today to discuss your predictive maintenance program. Email us at quotes@ivorbearings.com, call +1 305-769-9459, or visit our contact page HERE to get started.
The bearing gives you months of warning. A program is what lets you hear it.
