The engine already knows. The value is not prophecy — it is finding out on Tuesday in the yard instead of Saturday on a shoulder.
Key takeaways
Fault-code and telematics data can give you meaningful early warning on some failure modes — the engine control module is already monitoring sensors continuously and raising codes before a driver notices anything — but it does not predict a breakdown in any general sense, and no honest vendor claims a fleet-wide accuracy figure that travels across duty cycles. Geotab’s own documentation describes the mechanism precisely: engine faults are diagnostic trouble codes generated by the vehicle’s engine control module, which continuously monitors sensors and systems and generates a code when a potential issue is detected.
What that gives a small Canadian carrier is not prophecy. It is triage: a way to separate the code that can wait for the next scheduled service from the one that should pull the unit off a lane tonight.
Fault codes carry status and severity, and the distinction matters more than the code number. Geotab documents severity on a scale from Low, meaning no repair needed, to Critical, meaning check immediately. It also documents the status combinations: a fault that is Pending alone has not occurred often enough to be considered a serious issue and may be temporary, while a fault showing Pending and Active together has occurred repeatedly, is likely a real problem, is stored in the engine’s long-term memory, and means the vehicle should be inspected.
That is the whole predictive story in one paragraph, and it is genuinely useful. A code that keeps re-arming is a different object from a code that fired once on a cold morning. A fleet that treats them identically either ignores everything or chases everything.
1. Fault codes. Highest signal, lowest effort. The vehicle is telling you something it has already detected. Samsara’s developer documentation lists monitoring for fault codes, managing driver vehicle inspection reports and scheduling planned maintenance as the core maintenance capabilities of a telematics platform, which is a fair description of the category.
2. Usage-based intervals. Odometer and engine-hour feeds let you trigger service on what the unit did rather than on what you assumed it did. This is not prediction, it is accurate scheduling, and it delivers more avoided failures than anything labelled predictive.
3. Trend analysis. Fuel economy drifting on one unit, coolant temperature creeping up on a known grade, brake events per hundred kilometres climbing. Real, but it needs enough history and enough units to separate a trend from a season. A five-truck fleet in a Canadian winter has both a small sample and a large confound.
Be skeptical of three things. First, any accuracy or savings percentage quoted without a published method — ask what the denominator was and on what fleet. Second, the suggestion that predictive maintenance reduces your inspection obligations: it does not. Ontario still requires preventive maintenance inspections to sit inside a written maintenance plan performed by a qualified technician at prescribed intervals, still requires the periodic certificate inspection at a licensed station, and still requires the daily inspection. Third, the idea that a dashboard alert is a record. The compliance artifact is the work order and the certified corrective action, not the notification.
And the National Safety Code framework is unchanged by any of it. CCMTA’s Standard 11, Commercial Vehicle Maintenance and Periodic Inspection is the reference every province adopts or mirrors; telematics changes when you act, not what you owe.
Worked example: turning codes into a maintenance decision
The alert. A tractor raises an aftertreatment-related code on a Tuesday run. Status Pending only, severity low.
The first question. Has it re-armed? A single Pending code on one trip is not an intervention. The unit is flagged, not pulled.
Thursday. The same code returns Pending and Active together. That is the documented signal that the fault has occurred repeatedly and is stored in long-term memory. It now belongs on a work order.
The scheduling decision. The unit is due for a preventive service in three weeks and has a periodic inspection in seven. The service is pulled forward to this week; the inspection is left where it is. Two visits become one.
The record. The work order is written against the unit, the defect is coded, and the corrective action is certified. The alert itself is not the record and is not what an auditor will read.
What was not claimed. Nobody said the truck would have failed on Saturday. It might not have. The value is the avoided roadside event and the avoided second shop visit, both of which are real and neither of which needs a fabricated percentage.
No. The daily inspection is a separate legal duty on the driver and the carrier, and Ontario sets out what a defect means for whether the vehicle can move at all. Telematics has no standing in that determination.
Those are wear items, and wear is better handled by accurate usage-based intervals and by the trip inspection report than by fault codes. Treat brake and tire condition as an inspection question, and reserve fault-code work for systems the engine control module actually monitors.
Not necessarily. Electronic logging devices are certified against a national technical standard, and Transport Canada publishes the list of certified devices and accredited certification bodies. A vendor may sell both on one platform, but a maintenance feed is not a certified hours-of-service record.
The fault-code and usage-based layers usually are, because they cost little beyond the telematics you may already run for temperature monitoring or dispatch. The trend-analysis layer generally is not, because there are not enough units to see a trend before the season changes.
Predictive maintenance in Canadian trucking is mostly better-timed preventive maintenance. The engine already knows something is wrong; the win is that a person finds out on Tuesday in the yard instead of Saturday on a shoulder, and that the work lands on a scheduled visit rather than an emergency one. The judgement about whether a unit is safe to run stays with the technician and the driver, and the paperwork that proves it stays in the unit file.
A 30-minute call is enough to see how fault alerts, work orders and unit files could connect without another dashboard nobody reads.
