Why two tools give different figures from the same file
It unsettles people the first time and it rarely means either tool is wrong. They are reading
the same records and applying different assumptions — and the assumptions are where analysis
actually happens.
Three of them account for most discrepancies:
- How partial minutes are treated. Rounding conventions differ, and across a month
of activity changes they compound into visible gaps.
- What counts as the start of a day. Shift a day boundary by an hour and every
daily total in the period moves with it.
- How unrecorded time is handled. One tool leaves a hole; another assumes the
activity continued. Those are very different pictures of the same week.
Which is why comparing outputs is the wrong way to validate an analysis. The question isn't
which number is bigger, but which assumptions produced it — and a tool that won't show you its
working can't be checked at all.
The three errors that aren't errors
A large share of what an analysis flags as a problem is nothing of the sort, and knowing which is
which saves a great deal of wasted time:
- Events from a workshop visit. Power interruptions and sensor faults arise from
ordinary repairs. An event in the memory does not by itself prove tampering — but with
a dated workshop invoice beside it, it explains itself in ten seconds.
- Time discrepancies after a battery change. Real, harmless, and alarming to anyone
who doesn't know why they occur.
- Gaps at a change of employer or vehicle. Often a download that happened elsewhere
rather than time that went unrecorded.
What distinguishes an operation that handles this calmly from one caught out every time isn't the
software. It's a habit: every notable event gets a dated document, on the day. Six months
later nobody remembers why a unit briefly lost power on a Tuesday in March — and "I don't know" is the
only answer that genuinely costs money.
In short
Don't validate an analysis by comparing totals. Ask which assumptions produced them —
rounding, day boundaries, and how unrecorded time is treated. A tool that won't show its working can't be
checked.
The analysis gives you the question, not the answer
A misunderstanding that gives good tools a bad name: an analysis is expected to say what should be
done. It can't — and it shouldn't. It works with what was recorded and what has been
downloaded, and stops there.
Three things sit outside its reach: what was never recorded (driving with no card
shows as nothing, not as an error), what hasn't been downloaded yet (the best analysis
ends at the last download), and why any of it happened.
So the most useful figure isn't how many findings there were, but which ones repeat.
An isolated case speaks about one day. The same case on the same lane for a third month speaks about the
schedule — and only that second kind can be fixed at source.
This happens because new V2 tachographs use more advanced encryption protocols. If your analysis software is not updated to the 2026 version, it will be unable to decode the new GNSS geolocation fields. TachoTools is 100% adapted to this technology.