Telematics is what happens when a vehicle reports where it is and what it is doing to a computer somewhere else. A small unit behind the dashboard reads satellite positions and vehicle data, sends them over the mobile network, and a backend turns that stream into trips, stops, mileage and arrival times. Every delivery van, rental car and truck with a "track and trace" screen behind it runs on some version of this.
The part most people never see is how much the backend has to repair. A GPS position is a few metres off, it drifts while the van is parked, and between two pings the vehicle is invisible. We measured what that does to the two numbers fleets care about most: mileage and stops. Summing straight lines between GPS pings got the mileage of 10 simulated delivery and service rounds wrong by 16% with one ping a second and by 18% with one ping every 2 minutes, in opposite directions. Snapping the same pings to the road network with our Map Matching API brought the error down to 1.9% and 2.7%.
This article explains how a telematics system is built, how fleet tracking turns raw pings into routes, stops, mileage and ETAs, which EU rules apply, and what the experiment means if you build or buy one.
What is telematics?
The word joins telecommunications and informatics. In vehicles it means four parts working together:
- The tracking unit. A box (a telematics control unit, TCU) with a satellite navigation receiver, a cellular modem, usually an accelerometer, and a connection to the vehicle. It is either fitted by the car maker or added later.
- The vehicle data. Modern vehicles run their electronics over a CAN bus, a shared network between engine, gearbox, brakes and dashboard. A tracker reads it through the diagnostic port or a direct connection. Heavy trucks have a common interface for this: the FMS-Standard, agreed by European truck makers under ACEA, gives third-party telematics units access to vehicle data and remote download of the digital tachograph.
- Connectivity. The unit sends its data over the mobile network, usually in small batches to save power and data. When the network drops out, it stores positions and sends them later.
- The backend. Servers that receive the stream, clean it, store it, and turn it into something a dispatcher, a driver or an accountant can use.
A satellite position on its own is good to a few metres. The US government's GPS programme says GPS-enabled smartphones are typically accurate to within a 4.9 m radius under open sky, and worse near buildings, bridges and trees. A dedicated tracker with a good antenna does better, and a phone in a door pocket does worse. Either way, the backend never receives the road the vehicle drove on. It receives a cloud of dots.
How fleet tracking turns pings into routes, stops, mileage and ETAs
Everything a fleet platform shows is computed from those dots:
- Trips. Ignition on and ignition off, read from the vehicle or inferred from movement, cut the stream into trips.
- The route driven. Map matching snaps the noisy positions onto the road network and fills the gaps between them with the most likely path. This is the step that turns dots into a line on a street.
- Mileage. The length of the matched route, per trip, per day, per vehicle. It feeds fuel cards, tax logs, maintenance schedules and invoices.
- Stops. A run of positions that stays within a small radius for long enough is a stop. Stops are proof of delivery, billing time for a service visit, and the input for planning the next round.
- Geofence events. Entering or leaving a depot, a customer site or a city zone, computed by testing each position against stored polygons. Our guide to geofences covers how that test works.
- ETAs. The remaining route from the current matched position to the next stop, with its travel time, recalculated as the vehicle moves. Our live driver tracking tutorial builds this end to end.
Each step depends on the one before. A wrong match gives the wrong mileage, and a missed stop gives the wrong ETA for every stop after it. The single setting that shapes all of them is how often the tracker reports.
Our experiment: ping interval against mileage and stops
We built 10 realistic rounds with the MapAtlas Directions API on 2 October 2026: seven parcel rounds of 12 stops in Amsterdam, Berlin, Paris, Madrid, Milan, Warsaw and Vienna, and three service rounds of 5 or 6 longer visits around Lyon, the Munich region and the Utrecht region. Each round starts and ends at a depot on the edge of the area. Together they cover 534 km and 100 stops.
For each round we simulated a GPS tracker:
- Position error that drifts slowly, as real receivers do: 3 m per axis on average, with a correlation time of 60 seconds, plus 1% outlier fixes 30 m off, as from signal reflections between buildings.
- Time at each stop: 1.5 to 6 minutes on parcel rounds, 15 to 60 minutes on service rounds. The tracker keeps reporting while the vehicle is parked.
- Six ping intervals: 1, 5, 15, 30, 60 and 120 seconds, all sampled from the same simulated drive, so they differ only in how often they look.
Then we measured the mileage three ways (straight lines between pings, the same with parked pings dropped, and the route returned by the Map Matching API) and ran a standard stop detector on the pings.
| Ping interval | Pings per hour | Straight lines | Parked pings dropped | Map matched | Map matched, range per round | Stops missed (of 100) |
|---|---|---|---|---|---|---|
| 1 s | 3,600 | +15.7% | +8.2% | +1.9% | +0.3% to +4.7% | 1 |
| 5 s | 720 | +3.1% | +0.2% | +0.8% | -0.2% to +2.7% | 1 |
| 15 s | 240 | -1.4% | -3.0% | +0.3% | -0.8% to +1.7% | 1 |
| 30 s | 120 | -5.0% | -6.0% | 0.0% | -0.9% to +1.0% | 1 |
| 60 s | 60 | -10.6% | -11.3% | -0.9% | -3.7% to +1.9% | 1 |
| 120 s | 30 | -17.9% | -18.2% | -2.7% | -23.8% to +19.2% | 24 |
Errors are for the 534 km of all rounds together; the range column shows the best and worst single round. The one stop missed at 1 to 60 seconds is the same pair of Milan stops 13 m apart, which the detector reads as one stop.
Frequent pings count miles that were never driven
At one ping a second, straight lines between pings added 15.7% to the mileage, and almost all of it comes from standing still: in our model, as in real receivers, a parked van's reported position wanders by most of a metre every second, and every wobble adds distance. The service rounds, which spend most of their time parked, were worst: the Lyon round came out 35% too long.
Dropping pings that stay within 15 m of the previous one halves the problem, to 8.2%. Map matching removes most of it, to 1.9%, because a vehicle cannot drive back and forth along a kerb.
Sparse pings cut corners
At one ping every 2 minutes, a delivery van covers about 1 km in town and several kilometres on a ring road between two reports. Straight lines between those points cut every corner and every detour, and the mileage came out 17.9% short.
Map matching recovers most of it, to 2.7% short across all rounds, because it routes along real roads between the pings. Single rounds tell a less comfortable story. With 2 minutes between pings, the matcher has to guess the path, and the guesses ranged from 23.8% short in Vienna to 19.2% too long in Paris. At 60 seconds every round was within 3.7%, and at 15 to 30 seconds within 1.7%.
Stops disappear at 2 minutes
Up to 60 seconds, the stop detector found 99 of 100 stops. The one miss was a pair of Milan stops 13 m apart, which no interval separates: a backend sees two parcels next door as one stop.
At 2 minutes, 24 of the 84 parcel stops vanished. A 3-minute drop at a front door leaves at most one or two pings, and often both land while the van is still arriving or already leaving. All 16 service visits of 15 minutes or more were found at every interval.
What this means for teams that build or buy fleet tracking
- Map-match on the server. It turned a 16% overcount and an 18% undercount into errors of 2.7% or less, with the same pings. Mileage from raw GPS points is not usable for billing or tax logs at any interval we tested.
- Report every 5 to 30 seconds while moving. In our test this was the range where every round stayed within 2.7% after matching and the detector found 99 of 100 stops, at 120 to 720 pings an hour. Live tracking on a map or a customer-facing ETA may need more; mileage and stops do not.
- Do not report every second while parked. Either switch to a slow heartbeat when the ignition is off or filter parked pings before you add up distance. Parked jitter was the largest error source in the whole test.
- Two minutes is too sparse for doorstep deliveries. It still works for long service visits and for the rough mileage of a whole fleet, but it misses short stops and makes single trips unreliable.
- Keep the raw pings. A better matcher or stop rule next year can only be applied to data you stored.
What EU rules apply to telematics
GDPR, for the driver's location. The position of a company vehicle is personal data of the person driving it. The EU data protection authorities' Opinion 2/2017 on data processing at work (WP249) accepts vehicle tracking for legitimate purposes such as safety or locating vehicles, but says employers should not use it to track drivers' behaviour or whereabouts, that it is unlikely there is a legal basis for tracking outside agreed working hours, and that a driver allowed private use should in principle be able to switch location tracking off temporarily. In practice that means a stated purpose, a privacy switch, and retention limits in the product.
The smart tachograph, for professional drivers. Regulation (EU) No 165/2014 requires a tachograph that records driving and rest times in trucks over 3.5 t that fall under the EU driving-time rules. Its second-generation smart tachograph, amended by Regulation (EU) 2020/1054, records the vehicle's position at the start and end of the working day, every three hours of accumulated driving, at every border crossing and at loading and unloading. From 1 July 2026, vans over 2.5 t on international goods transport need one too. Telematics units often download the tachograph remotely, which is what the FMS interface above provides for.
eCall, in every new car model. Regulation (EU) 2015/758 requires an automatic emergency call system in new types of cars and light vans, type-approved since 31 March 2018. It is telematics with one job, and the regulation keeps it narrow: the system must not be traceable or subject to constant tracking, it may keep only the last three positions, and it sends only a minimum set of data when a crash triggers it. Delegated Regulation (EU) 2024/1180 moves eCall to 4G and 5G networks as 2G and 3G networks shut down, with the new standards applying from 1 January 2026.
The Data Act, for who gets the data. Since 12 September 2025 the EU Data Act gives users of connected products, vehicles included, a right to the data their use generates. Our Data Act explainer for location data covers what that means for vehicle positions.
What businesses use telematics for
- Delivery and logistics: proof of delivery from stops, live ETAs for customers, planned against actual routes for last-mile delivery rounds.
- Field service: time on site per visit, travel time between jobs, and the nearest technician to a new job.
- Mileage and tax logs: per-trip distance and private or business use, which many tax authorities accept from a tracker.
- Maintenance: service intervals from real mileage and engine hours, and fault codes from the CAN bus.
- Compliance: driving and rest times, speed against limits, and access to low emission zones.
- Insurance: usage-based policies priced on distance, time of day and driving style.
- Safety: crash detection, lone-worker alarms and stolen-vehicle recovery.
Fleets are one group of machines that read maps; our maps for machines overview covers the others, from robotaxis to drones.
Map-match a trace with the API
The map-matching step of this experiment is one request per trace. Send the pings with their timestamps, and the response contains the matched route as an encoded polyline, the road segments with their attributes, and a confidence score:
curl -X POST "https://gateway.mapmetrics-atlas.net/map-matching/?token=YOUR_TOKEN" \
-H "Content-Type: application/json" \
-d '{
"shape": [
{"lat": 52.35472, "lon": 4.89402, "time": 0},
{"lat": 52.35503, "lon": 4.89751, "time": 30},
{"lat": 52.35801, "lon": 4.90098, "time": 60}
],
"costing": "auto",
"shape_match": "map_snap"
}'
Measure the length of the returned shape for mileage. Where the matcher cannot connect two pings that are far apart, route between them with the Directions API instead of drawing a straight line, as we did.
Method and limits
- Routes: MapAtlas Directions API, costing
auto, 2 October 2026, no departure time, so no live traffic. Stops are seeded random points around each area's centre, snapped by the router to the nearest drivable road, visited in nearest-neighbour order. The routed length is the true mileage. - Simulated tracker: position error as a first-order Gauss-Markov process, 3 m per axis, 60 s correlation time, plus 1% outliers of 30 m. Real receivers vary by antenna, mounting and surroundings, and many apply their own filtering, for example freezing the position below a speed threshold. Constant speed within each road segment, no traffic lights or queues.
- Map matching: MapAtlas Map Matching API, costing
auto,shape_matchmap_snap, pings sent with timestamps in chunks of up to 1,500. Straight jumps of more than 500 m in the matched result, where the matcher could not connect two pings, were replaced by a Directions route between their ends. - Stops: consecutive pings within 50 m for at least 60 seconds; runs less than 2 minutes and 50 m apart merged. A true stop counts as found when a detected stop lies within 100 m and overlaps it in time.
- What this does not measure: ETA accuracy, live-tracking latency, data cost per vehicle, or cellular dead zones.
The script is in our research archive: scripts/seo/blog-keyword-research/what-is-telematics.mjs, with the raw results in docs/blog-keyword-research/what-is-telematics-2026-10-02.json.
Related
For the algorithm behind the matching step, read what is map matching. To put a live fleet on a map, follow the live driver tracking tutorial. The Map Matching API and Directions API used here run on EU-hosted infrastructure, so driver positions do not leave the EU on their way to the matcher.
Câu hỏi thường gặp
What is telematics?
Telematics is the combination of telecommunications and computing in a vehicle: a tracking unit reads the vehicle's position from satellite navigation and data from the vehicle's own electronics, and sends both over the mobile network to a server. The server turns that stream into trips, stops, mileage, driving time and arrival estimates. Fleet managers, insurers, car makers and emergency services all use it.
How does GPS fleet tracking work?
A tracker in the vehicle records a position every few seconds and sends it to a backend in batches. The backend snaps the positions to the road network (map matching), splits them into trips and stops, adds up the distance and compares the position with planned routes and geofences. In our simulation of 10 delivery and service rounds, a tracker reporting every 5 to 30 seconds gave mileage within 1% of the true distance after map matching.
How often should a fleet tracker send its position?
For mileage and stops, every 5 to 30 seconds while the vehicle moves is enough, if the backend map-matches the trace. In our test, 30-second pings still found 99 of 100 stops, and the map-matched mileage of every round was within 1% of the truth. At 2 minutes, 24 of 84 short parcel stops disappeared and single rounds were off by up to 24%. Live tracking on a map needs shorter intervals than mileage does.
Is fleet tracking legal under GDPR?
Yes, when the employer has a legal basis and the tracking is proportionate. The EU data protection authorities' Opinion 2/2017 on data processing at work accepts vehicle tracking for purposes such as safety and locating vehicles, but says employers should not use it to monitor drivers' behaviour or whereabouts, that there is unlikely to be a legal basis for tracking outside agreed working hours, and that drivers allowed private use should in principle be able to switch location tracking off.
