When Fleetpio shows a maintenance director that a pump has a health score of 71, the number needs to mean something actionable, not just "medium concern." The value of asset health scoring as a decision tool depends entirely on the maintenance team understanding what went into the number — and equally important, what it doesn't capture.
This piece is for maintenance directors and plant managers who are either evaluating condition monitoring systems or trying to get more out of a system they already have. We'll cover how health scores are typically derived, what the common failure modes of scoring systems are, and how to interpret scores in ways that improve maintenance prioritization rather than creating a new category of dashboard noise.
How a Health Score Is Actually Constructed
Health scores in condition monitoring systems are normalized composite metrics — they take multiple raw measurements, compare each against a baseline, weight them by significance, and roll them into a single 0-100 (or similar) number. The specific inputs and weights vary by system, but for rotating equipment the most common ingredients are:
- Vibration deviation from baseline: how far has the overall vibration level moved from the equipment's healthy state, across multiple frequency bands?
- Bearing defect frequency amplitude: are BPFO, BPFI, or other defect frequencies elevated relative to baseline?
- Temperature deviation: is the bearing housing or process temperature trending above the equipment's normal range at current load?
- Rate of change: is the trend accelerating? A score of 75 that has been stable for three weeks is different from a score of 75 that dropped from 90 in the past eight days.
In Fleetpio's system, the score is calculated daily for each asset and normalized to the asset's own baseline (established during the first 14 days of monitoring), not to a population average of similar equipment. This matters because "normal" vibration for a 15-year-old pump in chemical service is different from normal for the same model newly installed in clean water service. Normalizing to the asset's own fingerprint reduces false positives from equipment-to-equipment variation.
What Score Thresholds Should Trigger What Actions
The threshold bands we use as defaults, and the reasoning behind them:
Score 80-100: Healthy. Asset is operating within normal variation of its baseline. No action required beyond continuing to monitor. Trend stability matters here: a score that has sat at 85 for 60 days is more definitively healthy than one that recovered to 85 from a recent dip and is still fluctuating.
Score 65-79: Watch. Departure from baseline is detectable but not yet severe. This is where you look at which components drove the score down — is it a bearing frequency that appeared, a temperature trend, or broadband vibration elevation? The watch stage is the right time to increase collection frequency (if you're doing periodic rather than continuous collection) and to make sure the relevant spare parts are available. You're not scheduling maintenance yet, but you're preparing.
Score 40-64: Scheduled intervention. The degradation trend is clear. At this point, Fleetpio auto-creates a draft work order and pushes it to your CMMS. The work order contains the fault type classification, the supporting sensor data, and a recommended maintenance action based on the fault signature. The technician should be scheduled within the next 1-3 weeks, not deferred to the next quarterly PM cycle.
Score below 40: Urgent. Getting below 40 in normal Fleetpio deployment means something was missed in the watch/intervention stages. Equipment that is degrading steadily doesn't typically jump from 75 to 35 overnight — it moves through the watch stage. If you see a score below 40, check your alert delivery (were watch-stage alerts going to someone who was out?), check whether the equipment was recently added and the baseline is still calibrating, and schedule immediate inspection regardless of shift timing.
The Rate of Change Is as Important as the Score
Maintenance directors who check the score once a week and act on the absolute number often miss the most important information: how fast is the asset degrading? Two assets both at score 72 on Monday present very different risk profiles if one has been stable at 72 for two weeks and the other dropped from 89 to 72 in eight days.
The rate-of-change question matters most in the watch band (65-79). An asset that enters the watch band and remains there for weeks is probably degrading slowly — you have time to plan. An asset that drops through the watch band to 64 in 10 days is telling you the failure mechanism has an aggressive progression, and the scheduled intervention window needs to be tighter. In Fleetpio, we show a 30-day trend line alongside the current score for exactly this reason: the direction and velocity of the trend is the context that makes the number actionable.
What Health Scores Don't Capture
This is where we want to be direct about limitations, because over-trusting a health score is its own failure mode.
Health scores based on vibration and temperature do not reliably capture: sudden catastrophic failures (a seal that fails due to an abrupt process upset, a shaft that breaks due to a preexisting crack that wasn't visible in the vibration spectrum), lubrication failures in the early stage before thermal or vibration consequences appear, and external damage (corrosion, impact, cavitation from process conditions outside normal range that resolve before the next collection window).
We're not saying condition monitoring replaces human inspection. A reliability engineer who walks the floor monthly catches things that no sensor array captures — corrosion on pipe flanges, soft-foot on a pump baseplate that's contributing to misalignment, an unusual process condition that a technician knows is unusual because they remember last quarter. The right relationship between health scores and physical inspection is complementary: use scores to prioritize which assets need attention and to extend inspection intervals on assets that are consistently healthy, but maintain the physical inspection program.
A Scenario: Using Scores for Prioritization Across a Fleet
Consider a maintenance director responsible for 65 rotating assets across two production buildings. Without condition monitoring, the team runs through a prioritization process at every Monday morning shift-start meeting that involves asking floor technicians what they heard, checking whether anything alarmed over the weekend, and making judgment calls about which assets are highest risk going into the week.
With Fleetpio health scores updated each morning, the Monday meeting starts with the fleet sorted by current health score, with trend direction visible. The three assets in the watch band are reviewed in order of rate of change. If one dropped 12 points last week and another dropped 3 points, they prioritize the faster-degrading one for closer attention. Work orders for the two assets in the scheduled-intervention band are already in the CMMS from the previous night's auto-generation. The team can spend meeting time on the judgment calls — which downtime window works best for the scheduled replacements, whether to bring in a contract technician for the larger job — rather than on figuring out which assets to worry about.
That's the maintenance coordination value of health scoring: it replaces a semi-informal process with a daily data feed that imposes a consistent prioritization framework regardless of who's managing the Monday meeting.
Connecting Health Scores to Maintenance Decisions
One adoption barrier we see with maintenance directors new to health scoring is the question of accountability: if a work order is triggered by a system score rather than by a technician's recommendation, who owns the decision? This concern is real and worth addressing directly.
The answer is that the health score triggers a work order for review, not an automatic maintenance action. The work order created at score 65 goes into the CMMS as a draft. A maintenance planner or supervisor reviews the fault type, the supporting sensor data, and the recommended action before approving and scheduling. The system provides the early warning and the technical context; the maintenance team owns the decision. The score is a trigger for a structured review process, not a mandate.
What changes is the time available for that review. When a score-triggered work order arrives 3-4 weeks before expected failure, the maintenance team has time to source parts, pick an optimal maintenance window, and prepare the technician with the right information. When a failure is discovered through a weekend alarm or a technician noticing something wrong, the work order is created in an emergency context where all those options are compressed. Health scoring doesn't remove human judgment from maintenance decisions — it creates the lead time for that judgment to be exercised without the pressure of an imminent failure.