This is Part 1 of The Uptime Equation, a series on the economics of edge intelligence in commercial vehicles.
A heavy-duty truck comes into a dealer with an intermittent fault code. The technician replaces the most likely component, clears the code, and sends the truck back out. Three weeks later, it returns with the same code. Then it happens again. On the fourth visit, someone finds the real cause: a chafed wiring harness two modules away from where the code pointed. Across those visits, the fleet lost the truck for most of a week. The OEM paid for three parts that weren’t broken and three rounds of warranty labor, and it took a hit to a relationship it will need at the next fleet renewal.
Nobody did anything wrong. The fault code fired exactly as designed. A J1939 fault code (SPN/FMI) tells a technician which signal left its expected range and in what way. It doesn’t say why. A single wiring or connector issue can trip codes across several modules at once. Commercial vehicles have run telematics-driven maintenance for years, with DTC-triggered work orders, dashboards, and alerts, and that loop works. What’s limited is the resolution of the signal feeding it.
What missing context costs
According to ATRI’s 2026 Analysis of the Operational Costs of Trucking, the average cost to operate a truck reached $2.336 per mile in 2025, the highest ever recorded. Repair and maintenance rose 8.6% to $0.215 per mile, the largest increase of any line item except tolls. As repair costs climb, every repeat visit caused by an ambiguous code costs more. Slow diagnosis is common: McKinsey’s 2026 survey of aftersales and quality executives found it takes about seven weeks to reach a confirmed root cause, and cited poor data quality as a key barrier to using AI to predict quality issues.
For the OEM, that ambiguity shows up as warranty expense. Across the auto industry, warranty costs reached $58 billion in 2024, double their 2012 level, as noted in the Sonatus blog post The $58 Billion Problem Vehicle AI Is Built to Solve. Every part replaced under warranty unnecessarily and every “no fault found” return is part of that number.
Uptime helps decide whether a commercial line wins the next fleet contract, and fleets increasingly write uptime and diagnostic capability into their purchase requirements.
Missing context has a human cost, too. When alerts are vague, late, or wrong often enough, drivers stop acting on them and technicians start checking dashboards by hand. An alert nobody trusts is worse than no alert, because it trains people to ignore the one that matters.
Fragmentation, across the fleet and inside the truck
Across the fleet, the problem is inconsistency between brands. The Road Ahead: Unlocking New Opportunities with Standardized Commercial Vehicle Data, highlighted the fact that over 90% of fleets run mixed-brand vehicles, and proprietary signal formats and naming conventions make integration hard. COVESA projects the commercial vehicle telematics market to grow from $33 billion in 2022 to $95 billion by 2032, and all of that growth sits on this fragmented data layer.
Inside a single truck, the problem is ownership. Heavy-truck OEMs are also integrators. Engine, transmission, after-treatment, and brake controllers often come from different suppliers, and each reports faults in its own way. The OEM owns the vehicle and the warranty relationship, but not every ECU that emits a code. Standards like COVESA VSS help with naming across brands. They don’t, however, add the context that a supplier ECU’s code leaves out.
What vehicle context actually means
Vehicle context doesn’t mean more data. It means the signals, system state and history around an event, captured at the right moment:
- The window around the fault, not just the code. Signals from before and after the trigger show how the condition developed.
- Cross-domain state at that moment. That means what the engine, aftertreatment, electrical system and duty cycle were doing at the same time.
- Collection triggered by the event itself. Capture is targeted to the question being asked, rather than logging everything or sampling on a fixed schedule.
- History kept on the vehicle. When a pattern emerges, an engineer can look back weeks instead of waiting for the fault to recur.
This is the idea behind event-driven data collection such as Fastlane™ Collector, and it maps to the Collect stage of the Agentic Loop. It also changes what the data is good for. A fault code with no context can open a work order, but it’s poor material for training a predictive model, because a model trained on ambiguous labels learns the ambiguity. Context-rich event data is what makes AI/ML-based prediction workable. We’ll come back to that in a later installment of this series.
Why this gets decided on the vehicle
A better dashboard can’t recover context the vehicle never captured. Vehicle context is decided at the edge, on the vehicle in the seconds around an event, by what the platform is built to sense, keep, and send. That makes it an OEM architecture decision. Fleets feel the problem and push it upstream through their purchase requirements. The OEM’s platform and E/E architecture teams are the ones who can solve it.
Capturing that context costs something: sensing, onboard compute, storage, and bandwidth. OEMs manage bill of materials (BOM) cost across commercial lines with thinner margins and lower volumes than passenger vehicles, so this isn’t a casual decision. It comes down to two paths: the OEM can pay for more capable hardware to capture fidelity, or build software efficient enough to get it from the hardware already on the vehicle.
The bottom line: Commercial fleets already act on vehicle data. The gap is vehicle context: fault codes flag a monitored parameter, not a failed part, and that ambiguity drives repeat repairs, warranty cost, alert fatigue and training data too coarse for AI. Fixing it is an OEM platform decision made at the edge, on the vehicle at the moment of the event, and it forces a cost trade-off we take up in Part 2.