What
That disagreement is worth taking seriously rather than dismissing. When a vehicle health score slopes downward while the individual counters behind it look steady, two very different situations can produce the same picture. Either something mechanical is genuinely wearing, or the score is being shaped by how its rolling window is set up. Ruling out the second possibility first is usually faster, and it protects you from pulling a healthy vehicle off the road for the wrong reason.Start by mapping what actually feeds the score. A vehicle health score typically compresses monitored inputs such as fault codes, battery voltage, idle ratios, and mileage deltas into one number, but the exact input list varies by platform. Ask your vendor which inputs are included, which are weighted, and which are excluded. If idle ratio or battery voltage carries meaningful weight, a shift in duty cycle alone can pull the score down without any part failing. Confirm this in writing before you change thresholds or schedule inspections.Next, look at the rolling window itself. Health scores usually refresh on a rolling window, so freshness matters more than the value in front of you. If that window is longer than the span of the fault you are investigating, a repaired vehicle may still carry the old dip inside the average. The score is reporting history, not current state. Compare the window length against your repair timeline before concluding that a fix failed.Ignition-off data is a frequent culprit. When a parked battery voltage drops, averaging it into the same window as active driving data can make a perfectly healthy vehicle look degraded. Ask whether idle and parked periods sit in a separate bucket from driving data. If they do not, the apparent decline may be a configuration artifact rather than a battery on its way out. Keep in mind that separating these buckets is a useful step, not a guaranteed cure, so pair it with the other checks rather than stopping there.A score pinned at nearly the same value for weeks is a different signal again. Real mechanical behavior rarely stays perfectly still, so a flat score often points to a reporting or sensor issue rather than a stable vehicle. Check whether the reading is updating at all, and whether that asset class is even monitored the same way as the rest of the fleet. Trucks and light vans carry different fault signatures, and one shared threshold band across both will generate noise that looks like a trend.Once the configuration side is clear, turn to what a declining score can hide from the counters. Slowly wearing components, sensor drift, and intermittent wiring faults can all lower a score without leaving an obvious fault code behind. Some codes clear before anyone logs them, so the absence of a stored code is not proof that nothing happened. This is where the raw counters matter most. If your platform lets you retrieve them, compare them against the score movement directly rather than trusting the summary alone.A practical separation method helps here. Put score movement timestamps next to your own repair and fault records on one timeline. Then check depot-level patterns. If many vehicles move in the same direction at once, shared connectivity or environment explains correlated movement more often than simultaneous mechanical failure, so inspect the data path before pulling units from service. Test any threshold change on a single vehicle with a known-good baseline first, because fleet-wide changes hide the cause you are trying to find.Finally, escalate in writing. Document the discrepancy with dates, the inputs involved, and what you observed. Ask whether raw counters are retrievable, whether the health calculation runs on the device or the platform, and how sensor failures, wiring faults, and stuck readings appear in the score. Keep the software investigation and the physical inspection as separate workstreams, so neither masks the other. Remember that a clean score speaks only to monitored parameters, and unmonitored systems can still fail. Keep driver reports in the loop, and treat the disagreement as evidence to follow, not a number to argue about.Related FAQs: How can I tell if my ship's GPS is being spoofed?