Blog

Asset criticality: why predictive maintenance programs fail before they start

Programs stall when teams monitor the wrong assets. See how an asset criticality matrix shows where to start.

Date Published
November 13, 2026

Key Points

  • Predictive maintenance programs stall when teams monitor the wrong assets, so rank assets by criticality before you buy any hardware.
  • A cross-functional team of maintenance, operations, and safety can score about 50 assets in a half-day session.
  • Start with your highest-criticality asset: identify its most likely failure mode, then pick the technology that detects it earliest.
  • Customers in food and beverage, chemicals, and metals have reported savings from catching problems early.
  • AssetWatch reliability experts help your team turn what the data shows into planned maintenance action.

By Dustin Moore, Condition Monitoring Engineer Manager, and Kyle Privette, Lubrication Program Manager, AssetWatch

Why do so many programs stall?

The technology works. Vibration, temperature, oil analysis, and electrical monitoring all catch real problems early. The trouble usually starts before any sensor goes on a machine. Teams monitor whatever is easiest to reach, then discover that the failures that hurt the most still happen somewhere else. We cover the wider causes in why predictive maintenance fails.

“Some programs fail a year, year and a half into the program,” says Dustin Moore, Condition Monitoring Engineer Manager. “What we see is it’s not normally the sensors that cause the failure.”

Industry estimates put unplanned downtime at $1.4 trillion a year across industrial operations. Meanwhile, an often-cited figure holds that 80% of predictive maintenance programs stall within 12 to 18 months.

Rank Assets by Risk, Not Gut Feel

What do reliability teams say gets in the way?

Reliability professionals describe the same obstacles again and again:

  • Understaffing. The work backlog is bigger than the crew available to do it.
  • Output first. Operations won’t stop a running line just to take a look.
  • Tight budgets. Projects not tied to uptime or safety struggle to get approved.
  • Competing priorities. Five top priorities compete for the same quarter.
  • Burned before. Earlier systems overpromised or were never adopted.

None of these are arguments against predictive maintenance. They are arguments for making it smarter, and most trace back to one early decision. For more on getting crews on board, read how to improve predictive maintenance adoption across maintenance teams.

‍

Why do programs start in the wrong place?

Most programs begin with “What can we monitor?” That question spreads sensors across whatever is convenient. Low-impact assets get continuous coverage, while true line-stoppers stay reactive.

A better question is “Which assets would hurt the most if they failed without warning?” Answering that first gives you a plan that can be measured and defended.

“The most common failure mode we see in reliability programs isn’t the bad technology. It’s misallocated attention,” Moore says.

‍

How do you decide which assets matter most?

Start with an asset criticality matrix. A cross-functional team can score about 50 assets in a half-day session.

New to scoring assets? Here is what criticality analysis means.

Identify. List every asset and its function in the production process. Note location, redundancy, and single points of failure. Include the assets that run once a year.

Evaluate. Score each asset from 1 to 5 on five variables:

  1. Safety consequence. What is the worst-case injury or environmental impact if the asset fails?
  2. Production impact. Does failure stop the line, reduce throughput, or have no effect?
  3. Repair or replacement cost. What does the fix cost? Avoid over-weighting this factor.
  4. Failure probability. What is the historical failure rate, plus manufacturer data?
  5. Time to repair and redundancy. How long is the asset down, and is there a backup?

Prioritize. Sum each row and assign a tier:

ScoreTierReliability strategy
18–20Tier 1Continuous monitoring or prescriptive
16–17Tier 1Prescriptive to predictive
11–15Tier 2Predictive maintenance, route-based or continuous
8–10Tier 3Proactive or enhanced PM
Under 8Tier 4Run-to-fail, documented and deliberate

Replacement cost doesn’t equal criticality. A $200,000 pump with standby redundancy is less critical than a $4,000 seal on your only cooling water line.

Keep tier one small. “We should only be having that 20 to 25% being in that tier one,” Moore says. If most of the plant lands in tier one, the scoring needs another look.

‍

A worked example

The table below shows what a scored asset set can look like. Blower 1 carries a high safety score but stays in tier two because a standby unit covers it.

AssetSafetyProductionCostFailure probabilityRedundancyTotalTier
Motor 15545423Tier 1
Gearbox5544422Tier 1
Fan3544319Tier 1
Pump 13533317Tier 1
Blower 15522115Tier 2
Pump 23223212Tier 2
Pump 3221218Tier 3

‍

What are the most common pitfalls?

Three problems come up repeatedly.

  • Tier inflation. Everyone wants their assets in tier one. Set tier distribution targets before the scoring session begins.
  • Siloed scoring. Maintenance scores alone, and operations and safety aren’t in the room. Require cross-functional participation, because operations often surfaces what maintenance doesn’t know.
  • One-and-done. The matrix gets built once and never revisited. Set an annual review and an escalation path for surprise failures and revisit the matrix after any major equipment change.

‍

How do you match monitoring to each tier?

Criticality tells you whether to monitor and how often. Asset type and failure mode tell you which technology to use.

  • Tier 1 gets continuous online monitoring, typically 24/7 vibration and temperature.
  • Tier 2 gets route-based inspection, with scheduled vibration routes and enhanced PM tasks.
  • Tier 3 gets enhanced PM, with better lubrication intervals and inspection checklists.
  • Tier 4 runs to failure. Document the decision and replace the asset when it fails.

Vibration and temperature are the best starting point for most rotating equipment, including motors, pumps, fans, and compressors. Gearboxes are the main exception. Start gearboxes with oil analysis, then add vibration. Oil analysis catches contamination and lubricant breakdown long before vibration reacts.

Asset typeMost likely failure modeStart withAdd next
Electric motorBearing wear, imbalance, misalignmentVibrationElectrical monitoring
Centrifugal pumpBearing wear, cavitation, seal failureVibrationOil analysis if oil-lubricated
GearboxGear tooth wear, lubrication failureOil analysisVibration (spectral)
Electrical panelLoose connections, resistance heatingTemperature monitoringN/A
Slow-speed equipment (under 100 RPM)Bearing wear, lubrication deficiencyVibrationOil analysis

‍

Why does the P-F curve matter?

The P-F curve shows the interval between potential failure (P), when degradation becomes detectable, and functional failure (F), when the asset can no longer do its job. That interval is your detection window. The earlier you detect a fault, the more time you have to validate it, schedule it into a planned window, secure parts at normal cost, and coordinate with production.

Different technologies detect faults at different points. Oil analysis typically offers weeks to months of lead time. Spectral vibration typically offers weeks to months, while overall vibration level often gives only hours to days.

“Matching the technology to what we’re trying to catch comes into play here,” says Kyle Privette, Lubrication Program Manager.

Consider bearing wear, which progresses through four stages:

  • Stage 1, months out. The vibration spectrum still looks clean.
  • Stage 2, weeks out. Bearing defect frequencies appear in the spectrum. The overall vibration level still looks normal.
  • Stage 3, days to weeks. Defect frequencies grow with sidebands, and overall vibration begins rising.
  • Stage 4, hours to days. Overall vibration spikes, and operators can hear or feel the problem.

‍

What do real detection events look like?

Customers have reported savings from early detection in three domains. Each saving below is customer-reported, not independently verified.

  • Food and beverage. Temperature rise at an extruder pointed to a problem with the heat exchanger. The issue was found and resolved in Stages 2 to 3, before catastrophic failure. Customer-reported savings: $575,000.
  • Chemicals. Haystacking in vibration data indicated a lubrication deficiency in a centrifugal pump. The bearings were greased, which avoided a motor replacement and four hours of downtime. Customer-reported savings: $100,000.
  • Metals and processing. 1X turning-speed impacts in the time waveform pointed to coupling wear on a bridle roll drive. The motor-side coupling was found stripped and replaced. Customer-reported savings: $439,200.

Two of these three were caught at Stage 3. There was still time to act, but the window was days, not weeks.

‍

How do you make the business case?

When it comes to caring about the same program, different audiences have different reasons. Maintenance wants alerts it can trust. Operations wants fewer unplanned stops and no production disruption. Leadership wants to know the risk is low and the spend is worth it. See how to build predictive maintenance buy-in.

Keep the model simple enough to defend, and build it on one asset first:

  1. Identify one Tier 1 asset from your criticality matrix.
  2. Cost one unplanned failure. Multiply downtime hours by production rate, then add repair costs, overtime, contract labor, and any secondary damage.
  3. Compare that to the monitoring cost for that asset at that coverage level.
  4. Present protection, not spend. Frame the message as: “We spend $X a year to protect $Y in downtime risk per event.”

A single defensible number beats a program-wide estimate nobody believes. “You have to have that one that you can defend and show the value of first,” says Moore.

‍

Want to see the full criticality framework?

This post introduces asset criticality as the starting point for any predictive maintenance program. The on-demand recording of “Not All Assets Are Equal: Using Criticality to Build a Smarter PdM Program” walks through the full scoring process, including the five variables, the tier bands, and a worked example with a scored asset set. Moore and Privette also cover the pitfalls that derail matrices, from tier inflation to one-and-done reviews.

Watch the on-demand webinar

‍

Frequently asked questions

What is an asset criticality matrix?

An asset criticality matrix ranks each asset by how much it would hurt if it failed without warning. Teams score safety, production impact, repair cost, failure probability, and time to repair, then sort assets into tiers to decide what to monitor first.

‍

Who should be in the scoring session?

Maintenance, operations, and safety. Maintenance alone misses what operations knows, so a cross-functional group gives a more accurate score.

‍

How often should we review the matrix?

At least once a year, and after any major equipment change. Set an escalation path for surprise failures so the matrix stays current.

‍

What if we don’t have good failure history?

Start with the people who know the equipment. Bring operators, mechanics, and experienced technicians into the room and ask what problems they’ve seen on each critical asset. Their experience is often the best record available, and you can refine it as data accumulates.

‍

Do we need to buy monitoring technology before we build the matrix?

No. The criticality matrix is a planning exercise. It determines which assets deserve monitoring, how often, and which technology fits.

‍

Can an asset run to failure on purpose?

Yes, in some cases. Redundant, low-consequence, inexpensive-to-replace assets can be a deliberate choice. A documented matrix gives you a defensible basis for that call.

Stay Ahead of Equipment Issues

See how predictive maintenance can help your team prevent failures and keep critical assets running.

Upcoming Predictive Maintenance & Condition Monitoring Events

Discover the latest strategies in asset reliability and downtime reduction—browse our featured sessions below or view all upcoming events.

ISO Vibration Analysis Category III
Training
•
Nov 16, 2026
•
Dallas, TX
Machinery Lubrication I – Fort Myers, FL
Training
•
Nov 17, 2026
•
Fort Myers, FL
360 Training Summit
Training
•
Nov 17, 2026
•
Fort Myers, FL