On 11 June a truck did not move between 00:00 and 15:28. Odometer flat. Driver card pulled. In that window the toll system booked 309 kilometres for the same truck. Nobody cheated and the toll box was fine. The trip happened the afternoon before. The booking arrived at night.
That is where toll telematics reconciliation breaks in most fleets. The amount per kilometre is almost always right. The time is almost never right. Put both in one table and you see differences that do not exist. And you miss the ones that do.
The amount is big and visible. The German truck toll brought in around 12.96 billion euros in 2024. A Euro VI truck above 18 tonnes with five axles in CO2 class 1 pays 34.8 cents per kilometre. That is the tariff table of the German transport ministry, valid from 1 July 2024. Four components sit inside that rate: infrastructure, air, noise and CO2. In groupage, toll is 6.28 percent of the handling cost of a shipment. That is the DSLV cost index for the second half of 2025. Staff: 51.08 percent.
So fleets check the amount. Tariff class, axles, CO2 class, cents per kilometre. That is right, and we do it too. On 4 September our tariff class check ran over 5,512 German toll bookings from 52 trucks over 30 days. Result: zero class deviations. The amount was clean. (The check still found 3,228 euros. That is the potential of the next CO2 class, not an error. Another post.)
So the amount is rarely the problem. Still, fleet managers get a list every week that says "toll higher than telematics". The reason is not in any tariff table.
Toll Collect bills registered customers once a month. The individual trip statement comes as PDF and CSV. The CSV guide is from April 2026. It lists the columns: plate, date, start, booking number, entry, exit, kilometres, euros. And one sentence every reconciliation should know. The kilometres can differ from the odometer. Toll capture counts defined road sections, not driven kilometres.
EETS providers use another export. Every week we read the Toll4Europe export of a carrier with a little over 50 tolled trucks. It has 17 columns. Among them: transaction date, start time, end time, booking time, start date, end date. Five time fields for one trip. Which one the dashboard uses for the comparison decides the result.
Back to 11 June. The partner's dashboard attached the toll to the telematics shift by transaction date, to the hour. Result: a 94 percent gap between toll kilometres and driven kilometres. We compared the same truck by day instead. Plate plus calendar day plus route. Result: a 10 to 15 percent gap. That is the real gap. It consists of untolled urban kilometres, for example 14.6 kilometres between two districts of the same city. 94 percent became 15 percent. No record changed. Only the join.
The same pattern creates the famous weekend toll. A fleet runs Monday to Friday. The export shows bookings on Saturday. Nobody drove. The booking is on Saturday, the trip was on Friday.
The time axis has a second side. Telematics does not deliver trips either. It delivers positions. Our trip data run fetches them every hour for all vehicles and derives shifts. Over the last 60 days that was 1,452 runs, 24 a day, across 77 vehicle IDs. A shift needs an inserted driver card. No card, no shift. The truck drove, the table says zero.
On 23 July a partner's tour control flagged ten trucks with "toll higher than telematics". We checked the raw positions. Nine of ten had driven, 222 to 628 kilometres that day, with positions from 00:00 to 23:59. The kilometres were not missing in telematics. They were missing in the shift, because several trucks ran without a card for parts of the day. One truck was the real case: box silent from 20 to 27 July, one position a day, zero kilometres. The toll booked 739 kilometres for it on 23 July. That became the rule: toll kilometres are a floor. If telematics is below the floor, telematics is missing, not toll.
The third trap is the overnight trip. The shift starts in the afternoon and ends after midnight. The start day carries the telematics kilometres. The toll books part of it on the next day. Compared by day, it looks like two errors. Compared by tour, it is zero.
This is how we build the reconciliation today:
First: the key is plate plus calendar day in Berlin time plus route. Never the hour. Never the shift.
Second: compare net per kilometre. The target of 0.348 euros per kilometre for Euro VI, class 1, five axles is net. Gross is 0.414. A dashboard that shows gross produces a 19 percent phantom deviation. (We recalculated this in June on one truck. 18 bookings, 836.7 kilometres, 291.02 euros net. Result: 0.348. To the cent.)
Third: the floor rule. Toll kilometres less than or equal to telematics kilometres per day. Violations are data gaps, not toll errors. Each gets a reason: card missing, box silent, overnight trip.
Fourth: the driver comes from telematics, not from the toll export. On 12 August our weekly run assigned 1,238 of 1,352 bookings to a driver with confidence. 31 uncertain, 2 without GPS, 81 without a mapping. Those 114 lines sit in the report with a reason. They are the week's work for billing, the 1,238 are not.
The weekly run started twelve times in 60 days. Nine clean, three with a warning. One warning was an HTTP error in the driver match. The report went out in a resend run. The toll data was not the cause.
An honest caveat. This is operating data from one fleet and three cases. Whether another toll box sets its booking time the same way, we do not know. Only its export will tell. And the amount has cases of its own. The trip statement knows the types correction, credit and cancellation with fee. The CO2 class is a tariff attribute per vehicle. On 13 November 2025 the Bundestag added a review. Classes 2 and 3 get a check every six years. If the class on file is wrong, every kilometre costs too much. The tariff class check catches that. But it checks it apart from the time reconciliation. Two questions, two runs.
A weekly run that reads the toll export. An hourly run that turns positions into shifts. A table with the join by plate and day. A report with a reason for every line without a match. An overview that counts both runs. That is all. The effort sits in the reasons, not in the comparison. More on our logistics processes.
The toll box knows when the truck entered the motorway. The export does not always say so. As long as that holds, the calendar day is the best key.
Which time field does your reconciliation use today? If you do not know, send me one anonymised line of your toll export. I will show you which of the five times is the trip.
