The APM Layer as Reliability: Health, Predict, Monitor & AIP

🎯 Who this is for: Reliability engineers and asset managers deciding whether — and when — to layer Maximo's Asset Performance Management apps on top of a working Manage reliability program.

Series: Part 5 of 7 — MAS 9 Reliability Implementation Playbook | Read time: 18 minutes

🏔️ APM Sits On Top, Not In Place Of

There is a seductive sales narrative that goes: "MAS 9 has AI-powered Asset Performance Management, so buy Health, Predict, and Monitor and you have a reliability program." It is backwards. APM is an amplifier for a spine that already works — and if the spine is missing, APM amplifies nothing, because it reads from the same Manage objects (assets, work orders, meters, failure history) that Parts 3 and 4 built.

Asset Performance Management (APM) is IBM's umbrella for three reliability applications that sit on top of Manage. They are separately installable, and they all draw from the shared AppPoints licence pool — there is no separate per-app currency. In this series they belong to Phase 4 (Part 7): install them after the spine is populated and the core strategy is running, not before.

<aside>

💡 Key insight: Everything in the entire core reliability strategy — Failure Codes, criticality, Condition Monitoring, Meters, Job Plans, PMs — is stock Manage. You can run a complete, effective reliability program with zero APM add-ons. APM makes a working program sharper: better criticality, live anomaly detection, and failure forecasting. It does not make a broken program work.

</aside>

This part positions each APM app as reliability and shows how the data flows. It deliberately does not duplicate the deep application tours — the MAS-HEALTH, MAS-PREDICT, and MAS-MONITOR series each own their app in full. Here you learn what each one contributes to the reliability program and when to reach for it.

🩺 Maximo Health: Formal Scoring

Core Manage gives you an asset-priority ranking used as a criticality proxy (Part 3). Maximo Health is where a MAS 9 site gets formal scoring. It provides a consolidated view of asset health, criticality, and risk, computed by scoring "notebooks" — industry-standard formulas — that produce:

  • Health score — overall condition, from age, meters, failure history, and inspection data.
  • Criticality score — a formal, multi-factor score (safety, environmental, production, cost), not just a priority tier.
  • Risk score — condition × consequence.
  • End-of-life and effective-age scores — how much useful life remains.

The visualization surfaces

Health's value is as much in seeing the fleet as scoring it. Results surface through four view types:

ViewWhat it showsReliability use
TableRanked scores per assetTriage the worst actors
MapGeospatial colour-coded pinsSpatial risk concentrations (a bad substation, a corroded coastal line)
ChartHealth wheels, MTBF/downtime trendsTrend the program's effect over time
Matrixe.g. criticality × end-of-lifeFind risk concentrations to target

The criticality × end-of-life matrix

The matrix view is the workhorse. Plot criticality against end-of-life and the top-right quadrant — highly critical assets near end of life — is your capital and intervention priority list, computed rather than argued. This is the bridge into Asset Investment Planning below.

<aside>

💡 Key insight: Health's formal criticality score is the "upgrade" of the Part 3 asset-priority proxy — but the proxy is not wasted work. Health reads your Manage data (priority, meters, failure history) to compute its scores. A site that skipped the spine gets Health scores computed from garbage. The proxy is the input; the formal score is the output.

</aside>

📡 Maximo Monitor: "Is It Abnormal Now?"

Maximo Monitor is the IoT condition-monitoring platform. It ingests near-real-time sensor data, builds dashboards, and runs anomaly-detection rules across five rule families:

  • Threshold — a value crosses a fixed limit.
  • Statistical — a value deviates from its own historical distribution.
  • Spectral — frequency-domain analysis (vibration signatures).
  • Custom Python — arbitrary logic for domain-specific detection.
  • Generalised ML — learned anomaly detection.

Monitor is the real-time layer that the manual Condition Monitoring of Part 3 could only approximate with periodic readings. Where a technician logged a bearing temperature weekly, Monitor streams it continuously and flags the deviation the moment it appears — and that Monitor data also feeds Health scores and Predict models.

The distinction to hold onto: Monitor answers "is something abnormal right now?" It is a detector, not a forecaster. It tells you the bearing is running hot today; it does not tell you how many hours until it fails. That is Predict's job.

🔮 Maximo Predict: "When Will It Fail?"

Maximo Predict adds the machine-learning layer. Data scientists build models in Watson Studio / Watson Machine Learning, using IBM starter templates, in three model families:

Model familyTypical algorithmQuestion it answers
Failure probabilityRandom ForestHow likely is failure in the next window?
Remaining useful life ("days to failure")LSTMHow long until it fails?
Anomaly detectionIsolation ForestIs this pattern unusual?

Predictions surface in the shared Health/Predict UI. When an engineer reviews an at-risk asset, they can create a work order directly — which routes straight into Manage as a normal work order, planned and executed like any other. And a hard dependency: Predict requires Health. You cannot run Predict on its own; it builds on Health's data and scoring.

Monitor vs Predict — the crisp distinction

MonitorPredict
Question"Is it abnormal now?""When will it fail?"
MethodReal-time anomaly rulesTrained ML models
OutputAlert / dashboard flagFailure probability, RUL
AnalogySmoke detectorWeather forecast
FeedsHealth, PredictWork orders in Manage

<aside>

💡 Key insight: Monitor and Predict are complementary, not competing. Monitor detects the P on the P-F curve (Part 1) in real time; Predict estimates the slope to F. Together they turn "the bearing is warm" into "the bearing is warm and will fail in about 120 operating hours" — which is exactly the actionable signal that lets Manage schedule the repair before the failure.

</aside>

💰 Asset Investment Planning

New in MAS 9.1, Asset Investment Planning (AIP) takes the reliability conversation up a level — from "when do I maintain this asset?" to "when do I replace it?" It covers capital planning, investment strategies, and identifying the optimal time to replace an asset, and it builds on Health's risk and end-of-life scores to prioritise capital spend.

Two facts to plan around:

  • AIP requires the Maximo Optimizer component — it is not available on its own.
  • It is the natural consumer of Health's criticality × end-of-life matrix: the top-right quadrant becomes the capital-replacement candidate list, and AIP optimizes the timing of those replacements against budget.

This is where reliability meets finance. A reliability program that only ever schedules maintenance eventually hits assets that are cheaper to replace than to keep alive; AIP is the tool that turns Health's scores into a defensible capital plan.

🔄 The APM Data Flow

Here is how the whole layer sits on Manage and closes back into it:

 Manage (system of record: assets, WOs, meters, failure history)
   │
   ├── Monitor ──────► ingests live IoT / sensor streams
   │                    ("is it abnormal now?")
   │
   ├── Health ───────► pulls Manage + Monitor + geospatial/inspection
   │                    data → health / criticality / risk / end-of-life
   │
   ├── Predict ──────► runs ML on the combined data → failure forecasts
   │                    ("when will it fail?")  [requires Health]
   │
   └── AIP (9.1) ────► turns risk + end-of-life → capital replacement timing
                        [requires Optimizer]

   ALL of the above ──► create WORK ORDERS back in Manage
                         └──► planned & executed ──► failure reporting
                                └──► feeds Health/Predict ──► loop closes

Read the last two lines carefully, because they are the point: every APM app ultimately generates work orders back in Manage. APM does not replace the work-management engine — it feeds it better-targeted work. The loop closes when that work is executed, its failure reporting captured, and that data flows back up to sharpen the next round of scores and forecasts.

🧮 A Fully Worked APM Flow: One Critical Pump

Take pump P-101 from Parts 3–4 and watch the APM layer act on the spine you already built.

  1. Monitor streams the bearing-temperature and vibration signals that Part 3 only sampled periodically. A statistical rule flags that vibration is deviating from its own 90-day baseline — an anomaly now, three days before it would have crossed the action limit on the manual condition-monitoring point.
  2. Health ingests that Monitor signal plus P-101's Manage data (age, run-hours, failure history, the criticality rank from Part 3) and recomputes its scores: risk climbs, and P-101 moves into the top-right quadrant of the criticality × end-of-life matrix.
  3. Predict runs its LSTM remaining-useful-life model on the combined signal and estimates roughly 140 operating hours to functional failure — turning "abnormal" into "abnormal, and here is your window."
  4. The engineer reviews the at-risk asset in the Health/Predict UI and creates a work order directly. It lands in Manage as an ordinary work order, planned against the bearing job plan (Part 4), scheduled inside the 140-hour window.
  5. The loop closes: the technician executes the repair, reports the failure against the controlled vocabulary (Part 3), and that failure record flows back up to retrain Predict and re-score Health.

Notice what did the work: the spine. Monitor had a meter to watch, Health had criticality and history to weight, Predict had labelled failure data to learn from, and Manage had a job plan to execute. Strip the spine out and every step above degrades to noise.

<aside>

💡 Key insight: APM's payoff is measured in P-F interval terms (Part 1). Manual condition monitoring detects P at the action limit; Monitor detects it earlier from the trend; Predict estimates the slope to F. Each APM app buys you more of the warning window — which is exactly the lead time that lets Manage schedule the repair instead of reacting to the breakdown.

</aside>

🧩 Where Each APM Series Owns the Detail

To avoid duplicating the deep-dive series, here is the division of labour. This part tells you the reliability role; the named series own the how.

APM appIts reliability role (this part)Deep-dive series
Maximo HealthFormal criticality / risk / end-of-life scoringmas-health-series-index
Maximo MonitorReal-time anomaly detection ("abnormal now?")mas-monitor-series-index
Maximo PredictML failure forecasting ("when will it fail?")mas-predict-series-index
Asset Investment PlanningCapital replacement timing from risk scoresCovered within Health/Optimizer material
If you are configuring these apps, start with the deep-dive series. This part is for deciding whether and when they belong in your reliability program — and the answer is always "after the spine."

💳 Licensing & Packaging Facts

A few facts that decide whether APM is worth installing now:

  • Shared AppPoints. Health, Predict, and Monitor draw from the same AppPoints pool as the rest of the suite — there is no separate per-app currency, so the budget question is "do we have the points," not "do we buy another product."
  • HPU is gone. "Health & Predict – Utilities" (HPU) was discontinued as a standalone solution since MAS 8.11; its capabilities were folded into standard Health/Predict. Do not let anyone present HPU as a MAS 9 product — it is not one.
  • AIP needs Optimizer. Asset Investment Planning is 9.1-only and requires the Maximo Optimizer component. Budget for it explicitly if capital planning is on the roadmap.
  • Predict needs Health. A recurring scoping error is buying Predict without Health. Predict cannot run alone.

<aside>

⚠️ Watch out: The most expensive APM mistake is installing it to compensate for a missing spine. If your failure reporting is free text and your assets have no criticality, Health will score noise, Predict will train on garbage, and Monitor will alert on data nobody trusts. Fix the spine (Parts 3, 6, 7) first; then APM is a genuine multiplier.

</aside>

🚫 When NOT to Reach for APM Yet

  • The failure hierarchy is still free text. Cleanse it first (Part 6) — Health and Predict read from it.
  • No criticality ranking exists. Health upgrades a criticality proxy; it does not invent one from nothing useful.
  • Meters are sparse. Monitor and Predict are hungry for signal; without meter/sensor data they have little to work with.
  • The core strategy isn't running. APM sharpens decisions; if the decisions (Parts 1, 4) aren't made yet, there is nothing to sharpen.

📋 Practical Notes: APM Adoption Checklist

  • Confirm the spine is stable first — populated failure hierarchy, criticality on every in-scope asset, meters flowing (Parts 3, 6).
  • Install Health before Predict — Predict's dependency is not optional.
  • Start Health with the criticality × end-of-life matrix — it delivers the fastest, most defensible targeting value.
  • Pilot Monitor on a few sensored critical assets before a plant-wide rollout — prove the anomaly rules earn their alerts.
  • Route every APM-generated work order through normal Manage planning — APM's job is to create well-targeted work, not to bypass work management.
  • Budget AIP + Optimizer separately if capital replacement timing is a goal; do not assume it comes with Health.

Key Takeaways

  • APM sits on top of the Manage spine as an enhancement — it is Phase 4, not a prerequisite, and it reads the same objects the spine built.
  • Maximo Health gives formal criticality, risk, and end-of-life scoring via notebooks, surfaced in table, map, chart, and matrix views — the upgrade of the core-Manage priority proxy.
  • Monitor answers "is it abnormal now?"; Predict answers "when will it fail?" — they are complementary, and Predict requires Health.
  • All APM apps draw from the shared AppPoints pool and push work orders back into Manage, closing the reliability loop rather than replacing work management.
  • AIP (9.1, needs Maximo Optimizer) turns Health's risk and end-of-life scores into capital replacement timing; HPU was discontinued in 8.11 and is not a MAS 9 product.

References

IBM Official

Community

Series Navigation

Previous:Part 4 — From Analysis to Action: Job Plans, PMs & the Reliability Strategies App
Next:Part 6 — Populating the Spine: Data-Load Tools & Dependency-Correct Sequencing

About TheMaximoGuys: We help Maximo teams navigate the move to MAS 9 with practical, no-hype guidance grounded in how the platform actually behaves — from architecture and migration planning to the day-to-day work of configuring, extending, and running Maximo.

Published by TheMaximoGuys | July 2026