Fleet health and alerts

Every active vehicle carries a health score out of 100. It is not something you maintain: it is computed from the inspections, work orders, service schedules and maintenance costs already sitting against the vehicle. This is the answer to the question every fleet asks about its checklists — where does all this end up? It ends up here, weighted, and it is what raises alerts when a vehicle starts going wrong.

Half the score is inspections and servicing — the two things your team does every day. That is deliberate. A vehicle that is inspected regularly and serviced on time scores well even if it is old and expensive, because the fleet is doing the things that keep it safe.

Some of the detail is worth knowing, because it explains scores that look unfair:

• A defect that keeps coming back is penalised separately from the defect itself. The same item failing twice across the 90-day window counts as a repeat.
• Freshness counts. An inspection is worth full marks for a week, decays over the first month, and keeps decaying after that — so a vehicle that passed everything two months ago is not scoring as though it passed today.
• No inspections at all in 90 days scores a flat middling result, not zero and not full marks. So does having no active service schedules. The score treats missing evidence as unknown rather than good.
• Repair burden reads how long work takes, not just how much there is. Work orders that sit open for weeks cost more than the same work closed in two days.

The overall score is bucketed for reading at a glance, and the fleet health page shows the distribution across the whole fleet plus a filter per band:

• Excellent — 90 and above
• Good — 70 to 89
• Fair — 50 to 69
• Poor — 25 to 49
• Critical — below 25

The two lower bands are the ones wired to alerts, so the band boundaries at 50 and 25 are the numbers to watch rather than the score itself.

Four things trigger a recalculation:

• a scheduled overnight run that scores every active vehicle and then evaluates alerts;
• submitting an inspection;
• creating a work order;
• completing a work order.

The three event-driven ones run in the background rather than blocking your save, so a score checked immediately after completing a work order may still be the previous one. You can also recompute a single vehicle on demand from its health view.

Every run is stored, which is what gives each vehicle a score history and a trend. Nothing is recalculated retrospectively — a score is a record of what was known that day.

Each score carries a confidence level, which is a statement about your data rather than about the vehicle. It is high when the vehicle has recent inspections, a recorded mileage and a purchase date; medium when it has inspections or mileage but not the set; low when it has neither. A low-confidence score is a prompt to fill in the record, not to distrust the maintenance.

Trend needs history. It reads the direction of the last week of scores and reports improving, stable or declining — but only once there are at least seven daily scores to read. Until then it shows as unknown, so expect a new vehicle to have no trend for its first week.

Opening a vehicle’s health view breaks the overall score into the five sub-scores with the factors behind each one, alongside the score history. There is also an optional AI narrative that turns that breakdown into a written summary; it is rate limited and generated on request, not on every view.

An alert starts as new. Acknowledging it says someone has seen it, and only a new alert can be acknowledged. From there you either dismiss it — which is also available directly from new — or action it by attaching the work order raised in response. Actioning requires the alert to be acknowledged first, so the trail always reads seen, then acted on.

The two score-threshold alerts resolve themselves. A poor-health alert closes once the score climbs back to 50, and a critical-health alert closes at 25 — no cleanup needed once the underlying work is done. The other four stay until you dismiss or action them.

Alerts at critical and high severity also push a notification to every admin, which is why the severity assignments in the table above are worth reading before you turn scoring on for a large fleet.