ChecklistIndustrials

Predictive maintenance readiness checklist: assets, data and people before budget

A plant is ready for predictive maintenance on a given asset when the asset's failures matter, give measurable warning early enough to act, can be sensed with the right instruments, are recorded well enough in the CMMS to learn from, and have a named person who will plan work from an alert. This checklist scores those conditions asset by asset, tells you what a low score means, and shows when a simpler maintenance strategy is the better choice.

Reviewed 7 min read

On this page
  1. Assets where predictive maintenance is the wrong strategy
  2. The P-F interval test every candidate asset must pass
  3. Section A: asset criticality and failure modes
  4. Section B: sensing and the data path
  5. Section C: CMMS history and failure coding
  6. Section D: who acts on an alert
  7. Reading the scores section by section
  8. Scoring critical pumps and compressors at one site
  9. Readiness gaps teams tend to skip
  10. Questions and answers
  11. Sources

Assets where predictive maintenance is the wrong strategy

Run this filter before the checklist. Reliability-centered maintenance, whose minimum criteria are set out in SAE JA1011, chooses a strategy per failure mode rather than per plant1.

  • If

    Failure is cheap, safe and quick to recover from, such as a small redundant fan.

    Then

    Run to failure and keep a spare on the shelf.

    Monitoring would cost more than the failures it prevents.

  • If

    The failure mode is wear that tracks hours, cycles or throughput closely.

    Then

    Use usage-based replacement or inspection intervals.

    A counter predicts this failure as well as a model and is far simpler to run.

  • If

    The failure gives no detectable warning before it happens.

    Then

    Consider redesign, redundancy or a different component.

    Without a warning period there is nothing for a sensor or model to detect.

  • If

    The item is a protective device that only shows failure on demand, such as a relief valve.

    Then

    Schedule failure-finding tests at a justified interval.

    Hidden failures need proof testing, not trend analysis.

  • If

    Failure is costly or unsafe and degradation is measurable well before it occurs.

    Then

    Score the asset with the checklist below.

    This is the profile where condition-based and predictive approaches earn their cost.

The P-F interval test every candidate asset must pass

The P-F interval is the time between the point a developing failure becomes detectable and the point the asset fails functionally. Two conditions must hold for prediction to help: you must be able to check the condition more often than the interval, and the remaining time after detection must cover planning, parts and a maintenance window.

A bearing defect that shows in vibration spectra long before seizure usually passes. A brittle fracture that gives no measurable precursor does not. Agree the likely interval for each failure mode with the people who maintain the asset, then check it against your inspection or sampling frequency and your typical lead time for parts.

Section A: asset criticality and failure modes

Score every item zero (absent), one (partial) or two (in place) for each candidate asset.

0 of 5 checked

Section B: sensing and the data path

0 of 6 checked

Section C: CMMS history and failure coding

0 of 5 checked

Section D: who acts on an alert

0 of 5 checked

Reading the scores section by section

Read each section's share of the maximum available score; do not add sections into one total.

SectionMostly zerosMixedMostly twos
A: criticality and modesDo criticality and FMEA work before any purchaseFinish modes and intervals on the top assetsProceed to sensing design
B: sensing and data pathStart with portable routes or a short temporary installAdd permanent sensors to the top modes onlyReady for continuous monitoring
C: CMMS historyFix coding now; expect rules and anomaly detection firstClean recent records and code new ones properlyHistory can support supervised models
D: who actsDo not launch; alerts will be ignoredName owners and pilot on one areaReady to scale across assets

A low score in section D outweighs high scores elsewhere: monitoring without action is cost with no return.

Scoring critical pumps and compressors at one site

Readiness gaps teams tend to skip

Buying sensors before naming failure modes

Early signalDashboards full of trends nobody can interpret.

MitigationTie every sensor to a failure mode and an action in section A.

Treating the pilot asset as typical

Early signalA model that works on one machine fails on its siblings.

MitigationChoose pilots that represent a fleet of similar assets.

Ignoring the feedback loop

Early signalTechnicians close alerts without recording findings.

MitigationMake the finding a required field when the work order closes.

Questions and answers

How many recorded failures do we need before predictive maintenance makes sense?

There is no single threshold. Supervised failure prediction needs repeated, well-coded examples of each mode, which many plants lack for their most critical assets because those assets rarely fail. Anomaly detection, physics-based limits and expert rules work without failure labels, so a low failure count changes the starting technique rather than ruling out monitoring. Our model comparison covers that choice.

Should we fix CMMS data before or during a predictive maintenance pilot?

Start both together. Historic records can be cleaned for the pilot assets, but the larger gain is coding new work orders properly from day one, because those records will label future events. Waiting for a perfect history delays learning; ignoring coding means the pilot cannot prove anything.

Can portable vibration routes count as predictive maintenance readiness?

Yes, if the route interval is shorter than the warning period of the failure modes it covers. Route-based monitoring is often the sensible first step for assets with long warning periods, and it builds the analysis skills and records that a permanent system later needs. Permanent sensors make sense where the interval is short or access is unsafe.

What if our assessment scores low on who acts on alerts?

Treat it as the first thing to fix, not a detail. Agree who triages alerts, how they become planned work, and what happens when the planner disagrees. Pilot on one area with a reliability engineer who has time set aside. Technology spending before this is in place tends to end as unused dashboards.

Sources

  1. SAE JA1011_200908 Evaluation Criteria for Reliability-Centered Maintenance (RCM) Processes — SAE International · checked 10 October 2026
  2. IEC 60812:2018 Failure modes and effects analysis (FMEA and FMECA) — International Electrotechnical Commission · checked 10 October 2026
  3. ISO 17359:2018 Condition monitoring and diagnostics of machines: General guidelines — International Organization for Standardization · checked 10 October 2026
  4. ISO 20816-1:2016 Mechanical vibration: Measurement and evaluation of machine vibration, Part 1: General guidelines — International Organization for Standardization · checked 10 October 2026
  5. ISO 13374-1:2003 Condition monitoring and diagnostics of machines: Data processing, communication and presentation, Part 1: General guidelines — International Organization for Standardization · checked 10 October 2026

More in Industrials

Back to Industrials

Next step

Score your first candidate assets with a second pair of eyes

Send your asset criticality list, a sample of recent work orders and what you monitor today. We will return a scored view of the top candidates and the gaps to close first.

Send your asset list