Task · fleet

Fleet data analysis

One vehicle tells you what happened. A fleet tells you how often, under which conditions, and whether it is getting worse.

By Chayan Das, Founder & Lead Engineer

In short

Fleet data analysis is the analysis of measurement and diagnostic data collected from many vehicles rather than one, in order to find rare intermittent faults, quantify how often a condition occurs, and see degradation patterns before they reach the customer. It depends on batch-processing recordings consistently, matching signals across vehicles that do not name them identically, and querying results rather than opening files one by one. VehicleLab supports this locally: MDX indexes and batch-processes MDF4 logs through a Python SDK and REST API, and DiagAI applies the same diagnostic agents across the set, on your own infrastructure.

One DFC across the dropfleet_drop_2026_w38/

Occurrences per vehicle, segmented by drive phase — what turns "it happens sometimes" into a condition you can reproduce · illustrative

Fleet data analysis automotive teams actually run

A single recording proves a fault exists. A fleet answers whether it is a pattern: how many vehicles show it, which operating conditions precede it, whether it correlates with temperature, altitude, duty cycle or mileage, and whether the trend is worsening.

That changes the shape of the work. Instead of an engineer opening files, a defined analysis runs over every recording in a campaign and writes results you can query — counts, event tables, distributions and flags — with the underlying file always one click away when a case needs to be looked at properly.

Rare intermittent faultsEvents detected by rule across the whole set, including the ones no one was watching for.
Segmented by drive phaseCold start, warm idle, acceleration, cruise, deceleration and the key-off gap counted separately — the segmentation is usually what turns a vague fault into a reproducible one.
Condition correlationResidence distributions and operating-point statistics that show where the fleet actually runs.
Degradation over timeRepeat measurements compared across the campaign rather than in isolation.
Consistent namingCross-DBC and OEM signal matching so vehicles that label signals differently still compare.

Batch processing that stays reproducible

The analysis that matters is the one you can re-run in six months and get the same answer from. MDX exposes a documented local REST API and a Python SDK, so a campaign analysis is a script under version control rather than a sequence of clicks, and results land in a time-series store instead of a spreadsheet.

The same interfaces let AI agents operate the analysis over MCP — useful for triage across a large drop — while the recordings themselves stay where they are.

Two things make a fleet analysis survive contact with reality. Signals have to be matched across vehicles and software versions, because the same physical quantity is not always labelled identically, and invalid samples have to be excluded before any count or distribution is computed. Both are boring; both decide whether the numbers at the end mean anything.

Keeping fleet data where it belongs

Fleet recordings are among the most sensitive data an OEM holds: customer usage, locations, and the failure modes of a product in the field. The desktop instruments run fully offline, and the agent-based tools deploy on-premise with per-programme isolation, so the analysis can happen without the data moving.

The instruments

What does what.

Standards & references

Check the source.

These primary sources define the formats and regulatory context discussed above. They do not certify or endorse VehicleLab products.

Questions

Frequently asked.

What is fleet data analysis automotive engineering teams actually do?

It is the analysis of measurement and diagnostic data from many vehicles together — rather than one recording at a time — to find rare intermittent faults, quantify how often conditions occur, and detect degradation patterns across a population.

How do you analyse MDF4 logs across a whole fleet?

Run the same defined analysis over every recording as a batch job through an API or SDK, store the results in a queryable time-series store, and keep the original files accessible so any individual case can be opened and inspected properly.

Does fleet analysis require uploading data to a cloud?

No. MDX runs locally with a REST API on the machine itself, and DiagAI can be deployed on-premise, so fleet recordings — which carry customer usage and location data — never have to leave your infrastructure.

How are signals matched when vehicles name them differently?

MDX does cross-DBC and OEM signal matching, so recordings that label the same physical quantity differently can still be compared across the fleet.

How do you make an intermittent fault reproducible?

Segment every occurrence by drive phase — cold start, warm idle, acceleration, cruise, deceleration, key-off gap — and by ambient and thermal state. A fault that looks random across a fleet is usually concentrated in one of those bins, and that bin is the test case.

Can in-use performance ratios be checked across a fleet?

Yes, and it is one of the more useful fleet questions: the monitor counters recorded in field logs show whether monitors are actually running in customer driving, rather than only in the drive cycles used during development.

Next step

Open one of your own files with us.

Send a measurement, an A2L or an EHANDBOOK container. We will open it in the instruments on a call, with your engineers in the room.

Book a technical demo Read the docs