Preventive vs Predictive vs Condition-Based Maintenance: PM, PdM, and CBM in Maximo

Applies to: Maximo 7.6, MAS 8, and MAS 9

Three work types in Maximo get misused and nearly become interchangeable: preventive, predictive, and condition-based maintenance. Whether you're confident you know the distinction between them or you see them as interchangeable, once you try to generate and monitor them in Maximo you'll find it is much harder than a simple definition suggests. If you don't know the specifics, the confusion can cost real money when it drives the wrong investment or the wrong report to a regulator.

While all three are designed to prevent failures, they handle it in different ways. A preventive work order fires on a schedule. A condition-based work order fires when a measured condition crosses a limit. A predictive work order fires when a model forecasts a failure that hasn't happened yet. Same goal, three completely different triggers, and in Maximo they live in three different applications and get measured three different ways.

Here's the short version before we get into the why.

If you only remember one thing, remember which application generates the work order. Everything else follows from it.

Preventive Maintenance (PM)

Preventive maintenance is work you decide to do on a fixed interval, whether the asset needs it that day or not. Change the oil every 3 months. Inspect the conveyor every 500 operating hours. Calibrate the gauge every year. You are not reacting to anything the asset is telling you. You are following a plan you built ahead of time, on the bet that doing the work on a cadence keeps failures from happening.

The interval can be time-based (every 90 days) or usage-based (every 5,000 miles, every 250 run hours). Usage-based PMs are where the confusion usually starts, so let me kill that myth right now: a meter-based PM is not predictive. "Every 3 months or every 3,000 miles, whichever comes first" is still a predetermined rule. There is no model, no forecast, no analysis of the asset's actual health. You picked the interval in advance and the meter just tells you when you've hit it. Counting miles instead of months does not make it a prediction.

While many see PM as "old school," it has major benefits over PdM and CBM in many cases. When you find that a maintenance activity needs to be performed at a predictable interval based on usage, decay, and so on, then you can use that predictable interval to create a preventive maintenance schedule. If the maintenance labor and associated parts aren't very expensive, then a PM schedule can be very practical. Having a PM schedule can save you the cost of paying someone to do an inspection or a walk-down only to find the maintenance is needed anyway. It can also save you the cost of installing sensors and monitoring.

In my experience, PM is the backbone of every maintenance program for one simple reason: it is the work you can plan, schedule, resource, and prove you did. That last part matters more than people give it credit for, and we'll come back to it when we talk about regulators.

Condition-Based Maintenance (CBM)

Condition-based maintenance triggers when a measured condition crosses a limit you've set. You put a sensor or a manual reading point on the asset, define a threshold, and when a reading breaches that threshold, you generate the work.

Vibration on a pump climbs past the action limit. Bearing temperature crosses the line. Oil analysis comes back out of spec. The asset is telling you something is wrong right now, and you respond.

Here's the key word: now. CBM acts on the present state. The condition has already changed. You're not guessing about the future, you're reacting to a current reading that breached a rule. That reaction can be fast and it can save an asset, but it is a response to something that has already started, not a forecast of something that hasn't.

CBM is more efficient than blind PM because you only do the work when the condition calls for it, instead of on a calendar that may be too aggressive or too loose. But it depends entirely on having the right measurement points, the right limits, and a reading frequency tight enough to catch the change before it becomes a failure.

Before the Internet of Things (IoT) became commonplace, CBM was typically based on the results of an inspection by a person. In Maximo, when the person doing an inspection updated a condition meter, that could then trigger the Condition Monitoring application to generate a work order. Since IoT is now common, it is more efficient and cost effective to bring those readings into Maximo through IoT connectors, APIs, and Monitor.

While most people see the efficiency gains and cost savings from CBM, there is one less obvious benefit that I think is important, and it is part of the decision to use CBM over PM or PdM. CBM is well suited to new assets, new working conditions for current assets, a new runtime schedule, or assets whose usage varies, because varying usage changes when maintenance is needed but is not always easy to predict or schedule around. CBM lets us do the needed maintenance without requiring predictable conditions or usage.

Predictive Maintenance (PdM)

Predictive maintenance uses condition data plus analytics to forecast a failure before it happens, and gives you a lead time to act inside.

This is the part that separates PdM from CBM, and almost everybody blurs it. CBM says "the vibration is high right now, go look." PdM says "based on the trend in this vibration signature, the operating profile, and what we've seen on similar assets, this bearing is likely to fail in about three weeks, so schedule the replacement before then." One reacts to a threshold that's already been crossed. The other models a future state and buys you time.

That lead time is the entire value of PdM. You're not waiting for an alarm. You're acting on a prediction while the asset is still running fine, which means you can plan the work, order the parts, and schedule the downtime instead of scrambling.

PdM is the most powerful of the three and the hardest to do well. It needs sensor data, it needs enough failure history or a good physics-based model to train on, and it needs people who trust the forecast enough to act on it before anything looks visibly wrong. That trust is usually the real bottleneck, not the technology.

An overlooked requirement for the prediction model to provide enough lead time is a somewhat predictable usage pattern. If the use of the asset is random, then you'll want to defer to CBM.

What Actually Makes Them Different

Strip away the marketing and the three approaches line up on a single continuum, defined by how much the asset's condition drives the decision.

PM is blind to condition. You act on a cadence, full stop. CBM reacts to current condition. You act when a present reading exceeds a limit. PdM forecasts future condition. You act on a prediction of where the asset is headed.

That continuum is the cleanest way to think about it, and it exposes the two myths that cause the most trouble.

Myth one: meter-based PM is predictive. It isn't. A usage-based interval is still a fixed rule you set in advance. No forecast, no model. The trigger is accumulated usage, not a prediction.

Myth two: condition-based is predictive. Also no. CBM acts on a threshold that has already been crossed. PdM acts on a forecast of a threshold that has not been crossed yet. The difference is tense. CBM is present tense, PdM is future tense.

So why does everyone mix them up? Because the underlying mechanics overlap, especially in Maximo. Meters feed both meter-based PMs and condition monitoring points (although they use different meters). Condition monitoring can be wired to trigger a PM. Sensor data feeds both condition monitoring and predictive models. The plumbing is shared, so people assume the concepts are interchangeable. They're not. The mechanism is the same in places. The decision logic is completely different.

Why the Distinction Matters for Regulatory Governance

This is where getting the terms right stops being academic.

Across regulated industries, the maintenance regulators actually audit is the one you can document and prove you completed on schedule. Pharmaceutical GxP environments, power generation under NERC, transit under FTA oversight, nuclear under NRC, process safety under OSHA PSM: they all share a common expectation. You defined a maintenance program, and you can show auditable records that the scheduled work was completed, by qualified people, on time, with the results recorded.

That expectation generally lands squarely on preventive maintenance. PM is the regulated backbone because it is schedulable and provable. You can pull a report that says these tasks were due, these were completed, here are the dates and the labor records and the sign-offs. Auditors love that because it is binary and traceable.

Condition-based and predictive maintenance are powerful for reliability, and regulators increasingly accept condition-derived or reliability-centered intervals. But here's the thing: the compliance burden still lands on proving you executed your defined program. If your program says a condition-based task gets done when a limit is breached, you have to prove the limit was monitored and the response happened. If a predictive model drives the interval, you still have to document the program and its execution. The regulator is not auditing your algorithm. They're auditing whether you did what your program says you do.

This is exactly why PM Compliance is reported as its own number, separate from a broader schedule compliance figure. Schedule compliance tells you whether you completed the work you planned for the period. PM Compliance carves out the regulated, auditable subset: did the scheduled preventive work get done on time. When an auditor walks in, that's the number they want, and that's the number that protects you.

If you take one governance lesson from this, take this one: never let fancier maintenance strategies erode the documented PM program. CBM and PdM should reduce the failures you suffer. They should not become an excuse to skip the scheduled, provable work the regulator expects to see.

How Maximo Implements Each

Maximo separates these three for a reason, and once you see where each one lives, the distinction stops being abstract.

Preventive maintenance lives in the Preventive Maintenance application. A PM record carries the frequency, either a time-based interval (FREQUENCY and FREQUNIT in days, weeks, months, or years) or a meter-based frequency tied to the asset's meters. When the PM comes due, Maximo generates a Work Order, and that Work Order carries the PMNUM that ties it back to its source. You measure this with PM Compliance: of the PM Work Orders that were due in the period, how many were completed by their target completion date. If you build your PMs from a Master PM, you manage frequency and Job Plan changes once at the Master PM and propagate to every associated PM, instead of editing them one at a time.

Condition-based maintenance lives in the Condition Monitoring application. You define measurement points on assets or locations, set warning and action limits, and record readings against them, either from sensors or manual observations. When a reading crosses an action limit, Maximo can generate work or trigger an associated PM. The trigger is a present reading breaching a limit. You measure this with response time: from the moment the limit was breached to the moment someone acted. Speed is the metric, because the value of CBM is catching the condition before it becomes a failure.

Predictive maintenance lives in Maximo Predict, part of the Maximo Application Suite, supported by Maximo Monitor for sensor ingestion and anomaly detection and Maximo Health for asset health scoring. Predict applies models to operating and sensor data to forecast likely failures and the timeframe around them. The trigger is a forecast of a future state, and the value is the lead time it gives you to plan the work before anything fails. You measure this with prediction effectiveness: were the forecasts accurate, and did you act on the valid ones inside the lead window.

Notice that each strategy has its own success metric, and they don't substitute for each other. PM Compliance answers "did we do the scheduled work." Condition Response answers "did we react to the alarm fast enough." Prediction effectiveness answers "did we act on the forecast in time, and did that forecast give us enough time to prevent the failure." Reporting one when you mean another is how programs end up measuring the wrong thing and congratulating themselves for it. If you want the deeper version of this, I wrote a whole piece on why PdM Compliance always trails PM Compliance.

Where This Leaves You

If you're sorting out your own program, start by labeling every maintenance strategy you run by its trigger, not by whatever the last vendor called it. Time or usage elapsed is preventive. A current reading breaching a limit is condition-based. A forecast of a future failure is predictive. Get those labels right and three things fall into place: you'll put the work in the right Maximo application, you'll measure each one with the metric that actually fits it, and you'll know exactly which number to hand an auditor.

Most organizations should get PM solid and provable first, layer condition monitoring onto the assets where readings give you real warning, and reserve predictive for the critical assets where the data and the failure consequences justify the investment. That order matches both the difficulty and the payoff.

If you want to put numbers behind the preventive side, I built a free PM Compliance Calculator that runs your scheduled program: plug in the PM Work Orders that were due and the ones completed on time, and it shows you where you stand. The full Maximo implementation, including how PM, condition monitoring, and predictive measurement connect, is in the Maximo KPI Guide to PM Compliance.

Frequently Asked Questions

What is the difference between preventive and predictive maintenance?

Preventive maintenance is performed on a fixed schedule based on time or usage, regardless of the asset's condition. Predictive maintenance uses condition data and analytics to forecast a future failure and act within a lead time before it happens. Preventive follows a cadence you set in advance. Predictive responds to a forecast.

Is condition-based maintenance the same as predictive maintenance?

No. Condition-based maintenance acts when a measured condition crosses a threshold that has already been breached, so it reacts to the present state. Predictive maintenance acts on a model's forecast of a failure that has not happened yet, so it works from the predicted future state. The difference is tense: condition-based is present, predictive is future.

Is meter-based PM considered predictive maintenance?

No. A meter-based PM, such as service every 5,000 miles or every 250 run hours, is still preventive. The interval is a fixed rule set in advance, and the meter only tells you when you have reached it. There is no forecast or model involved, so it is not predictive.

Which type of maintenance do regulators require?

Regulators generally require a documented maintenance program with auditable proof that scheduled work was completed on time. That obligation lands on preventive maintenance because it is schedulable and provable. Condition-based and predictive strategies are accepted, but you still have to document your program and prove you executed it.

Where does Maximo handle predictive maintenance?

Predictive maintenance lives in Maximo Predict, part of the Maximo Application Suite, supported by Maximo Monitor for sensor data and anomaly detection and Maximo Health for asset health scores. Predict uses models to forecast likely failures and their timeframes.

Can Maximo do condition-based maintenance without Maximo Predict?

Yes. Condition-based maintenance runs in the Condition Monitoring application in Maximo Manage. You define measurement points with action limits, record readings, and Maximo generates work or triggers a PM when a limit is breached. That is condition-based, not predictive, and it does not require Predict. However, using Monitor is recommended.

How do I measure each type of maintenance in Maximo?

Measure preventive maintenance with PM Compliance, the percentage of scheduled PM Work Orders completed by their target date. Measure condition-based maintenance with response time from a breached limit to action. Measure predictive maintenance with prediction effectiveness, meaning forecast accuracy and whether you acted within the lead window.

Should I implement PM, CBM, or PdM first?

Get preventive maintenance solid and provable first, since it is the regulated backbone and the easiest to plan and document. Add condition monitoring on assets where readings give you meaningful warning. Reserve predictive maintenance for critical assets where the available data and the cost of failure justify the investment.

Quick Maximo questions are always free. Reach out on LinkedIn. I never charge for chatting.