Vehicle data analytics is the practice of taking the signals a connected car produces, processing them, and turning them into a decision an OEM can act on. The signals cover almost everything a vehicle reports about itself and its use: battery state, fault codes, location, feature activation, service uptake, data plan renewals.
Almost every manufacturer now collects this. Far fewer have decided what each type of analysis is actually for, or which questions justify processing data in seconds rather than overnight.
That decision matters more since 12 September 2025, when the main obligations of the EU Data Act became applicable and the European Commission published its accompanying guidance for the automotive sector. The guidance separates raw and pre-processed vehicle data, which users can request access to, from inferred or derived data such as driving scores and predictive outputs, which sits outside the mandatory sharing regime. In other words, the analysis a manufacturer performs on its own data is now one of the few parts of the stack that remains its own.
The four types of vehicle data analytics
Most confusion about analytics programmes comes from mixing these four together in a single request. They answer different questions and they cost very different amounts to run.
Descriptive and diagnostic
Descriptive analysis answers what happened. How many vehicles in a market activated their connectivity in the first week of ownership? How many renewed after thirteen months?
Diagnostic analysis answers why it happened. Activation fell in one market after a change to the handover process at dealer level, or after a tariff switch. This is the layer most teams already have, usually in the form of a monthly report, and it is the layer that tends to arrive too late to change anything.
Predictive and prescriptive
Predictive analysis estimates what happens next. A pattern of readings across thousands of vehicles suggests a component is likely to fail inside a given mileage band. A drop in feature use in the second month suggests a cohort will not renew.
Prescriptive analysis goes one step further and recommends the action. Order the part into the dealer network ahead of the service visit. Trigger a targeted offer before the renewal window closes. The value sits in the recommendation, but only if there is a delivery path for it, which usually means an OTA channel or an in-vehicle touchpoint that can carry the message.
Real-time processing against batch processing
Real-time processing handles events as they arrive, within seconds. It suits decisions that expire quickly: a driver standing at a forecourt waiting for a payment to authorise, a vehicle throwing a fault mid-journey, a roaming device that needs a profile switch on crossing a border.
Batch processing collects data over hours or days and runs on a schedule. It suits questions about a population rather than a vehicle: which model year has the highest second-year renewal rate, which markets under-perform on feature adoption.
Real-time costs more to build and more to run. The useful test is how long the answer stays useful. If a decision can wait until tomorrow morning without losing value, batch is the correct choice and the saving is real.
Where vehicle data analytics earns its place
Predictive maintenance
Predictive maintenance is the application most manufacturers reach for first, because the return is measurable in warranty spend and unplanned downtime rather than in engagement metrics.
It also has the strictest requirements. A model needs enough vehicles reporting the same parameters consistently across model years and regions, which is a connectivity problem before it is a data science problem. Fleet operators feel this more sharply than private owners, since a vehicle off the road is revenue lost that day.
Worth noting for planning purposes: the Commission’s guidance treats predictive maintenance as a related service under the Data Act where it involves two-way digital data exchange, which brings it inside Chapter II obligations.
Commercial signals in usage data
Usage data shows which features hold attention after the trial period ends, which markets convert, and where activation drops off between build and first drive. Explore3 surfaces those adoption signals for manufacturers who want to see behaviour rather than assume it.
The platform underneath
None of the above works without a layer that ingests data from vehicles across different model years, regions and network operators, and normalises it into something comparable. Cubic3 Cloud handles that ingestion layer, alongside data residency requirements that differ by market.
Vehicle data analytics: common questions
What is vehicle data analytics?
It is the collection and processing of signals generated by a connected vehicle to produce a decision a manufacturer, fleet operator, or service provider can act on. It covers descriptive, diagnostic, predictive and prescriptive analysis.
What is the difference between real-time and batch vehicle data processing?
Real-time processing handles each event within seconds of it occurring and suits decisions that lose value quickly. Batch processing runs on a schedule against accumulated data and suits questions about groups of vehicles over time. Batch is substantially cheaper to operate.
Does the EU Data Act require manufacturers to share their analytics output?
No. The Commission’s guidance places inferred and derived data, including predictive analytics output and driver scores, outside the scope of mandatory sharing. Raw and pre-processed data that is readily available does fall within scope.
