STATE OF DRIVER RELIABILITY

The share of reviews mentioning an abandoned truck went from 18.8% in 2018 to 40.4% in 2025. That is the headline, and it is the one figure on this site that has moved in the same direction almost every year.

Almost. In 2021 it fell, from 26.6% to 23.9% — the only year in eight that went the other way, and the only year the climb paused before resuming at 30.1% the year after. We can see that the dip happened. What we cannot see is why, because nothing in a review says what the freight market was doing that month.

Verdict language rose nearly as fast, from 19.6% to 37.2%: carriers increasingly end a review with advice rather than a description. Going unreachable roughly doubled, 4.7% to 9.6%. Paid travel with no arrival went 1.7% to 5.7% — small in absolute terms, more than tripled in relative ones.

One thing fell clearly. Mentions of a cash advance dropped from 5.8% to 2.4%, the largest sustained decline in the set. Drugs or alcohol, late deliveries and logbook problems each drifted down by one to two points, which is close enough to flat that we would not call it a trend.

Every figure here is a share inside its own year. The number of reviews grows every year, so raw counts between years compare nothing at all.

If two things were unrelated, the share of reviews mentioning both would be the product of their separate shares. Twelve pairs beat that by a wide enough margin to be worth naming. The multiplier below is how many times more often the pair appears than chance alone would give.

The strongest is not close: paid travel with no arrival and a no-show appear together 5.1 times more often than chance. That is one story told twice — the carrier bought a ticket and nobody came — and it is why those two mentions should never be treated as two independent warnings.

The rest are more useful precisely because they are not the same event told twice. Drugs or alcohol with a no-show at 2.7×. Drugs or alcohol with paid travel at 2.7×. Late deliveries with refused loads and low miles at 2.5×. Theft or money games with equipment never returned at 2.3×. An abandoned truck with a loaded trailer at 2.1×.

Then a longer tail, all still above chance: paid travel with going unreachable at 1.9×, late deliveries with logbook problems at 1.8×, a cash advance with late deliveries at 1.8×, theft with logbook problems at 1.7×, aggression with refused loads at 1.6×, going unreachable with a no-show at 1.6×.

On a reference call this changes one thing. When the last carrier mentions one half of a pair, the other half is worth asking about by name — not because it follows, but because it turns up alongside more often than a coin would explain.

The pair that matters most here is the one that almost never occurs. An abandoned truck and paid travel with no arrival land in the same review 0.25 times as often as chance would give — four times rarer than if the two had nothing to do with each other.

Separately, neither is rare. An abandoned truck is mentioned in 33.5% of reviews, paid travel with no arrival in 6.0%. Together, almost never.

The reading is that these are two different people. One takes the assignment and walks away from it partway through, usually with the truck somewhere it should not be. The other never starts: the ticket is bought, the seat is booked, the orientation date passes. A screen built to catch the first will not catch the second, because the second leaves no gap in a work history and no equipment anywhere.

The same shape shows up more weakly elsewhere. An abandoned truck with drugs or alcohol runs at 0.31×, and drugs or alcohol with a crash at 0.47×. We report those as separate populations, not as evidence that one behaviour prevents another.

32.2% of reviews contain verdict language — stay away, do not waste your time, would not rehire. That share has climbed from 19.6% in 2018 to 37.2% in 2025, faster than most of the behaviour categories.

We count it, and we keep it out of everything else. A verdict describes the person writing the review, not the person it is about: it tells you how angry the last carrier was, not what the driver did. Mixing it into a behaviour category would let strength of feeling inflate a figure about conduct.

It is worth knowing on its own terms. A rising share of verdicts across eight years says carriers are writing to warn each other more than they used to. Whether that reflects worse behaviour or a stronger habit of writing it down, this data cannot separate, and we are not going to guess.

Seasonality was the obvious hypothesis, and it did not survive. Across eight years, the share of reviews mentioning an abandoned truck ranges from 30.5% to 34.9% depending on the calendar month — a spread of 4.4 points across the whole set. Paid travel spans 2.3 points, no-shows 2.0, aggression 2.2. Those are the sizes you get from noise, not from a season. There is no hiring month that is safer than another, at least not one this data can see.

Two other things get asked about, and the honest answer is different: we cannot check them at all. Geography is not in the data. Neither is tenure, or fleet size, or pay. The review carries who it is about, a rating, the text, and the date it was written. That is the whole file.

The distinction matters. Checked and found nothing is a result. Cannot be checked is a limit, and a site that blurs the two is one number away from making something up.

The annual figures above are drawn to the last complete year. This section is the running one: the share of each month’s reviews that mention each category, from January through 27 August 2026.

A month is a small slice, and small slices wander. There are 56 readings in the table below, and at the usual two-sigma bar chance alone would hand us about three findings a year that are not there. So the bar here is three sigma for the share in question: roughly eight points for a share near a third, four for one near 8%.

Exactly one reading in 2026 clears it. Seven of these eight months have no finding, and they are published anyway — a monthly report that discovers a trend every month is not measuring, it is writing.