Fleet telematics data integration: the trucks are instrumented, the decisions aren't

Every truck in a mid-size fleet reports each hard-braking event to a scorecard nobody at the carrier had to build. Ask the same fleet what a mile cost last month and roughly one in four isn’t tracking the answer.

That gap is the finding underneath a report the American Transportation Research Institute released on 9 September 2026, and it isn’t a sensor problem. The trucks are instrumented. Fleet telematics data integration — the pipe from the device into the systems where money and maintenance decisions get made — is what’s missing, and ATRI’s own numbers show the gap widening with every system a metric has to cross.

TL;DR: ATRI surveyed 111 motor carriers (median fleet 92 trucks) on 44 telematics KPIs. Safety metrics a telematics device computes by itself — critical events, hard braking, speeding, harsh cornering — are used by more than 90% of fleets. Cost per mile, which carriers rated the single most important operations KPI, is used by 73%. Vehicle downtime prediction, 49.5%. The pattern isn’t importance and it isn’t hardware: 48% of fleets don’t connect telematics to their maintenance software, 39% don’t connect it to billing and accounting, and 21% still dispatch by phone and text. The KPIs that changed a decision are the ones that only needed one system. Everything else is waiting on an integration nobody has built.

What ATRI actually measured

The report — From Data to Decisions: Trucking’s Use of Telematics Data to Improve Performance, by Jeffrey Short, Andrew Fain and Jack Spalding — asked carriers to rate 44 KPIs from “don’t use” to “critical,” then split the answers into two numbers: how important a KPI is to the fleets that use it, and what share of fleets use it at all.

Safety KPIs averaged 3.96 out of 5 on importance. Operations averaged 3.35. Maintenance, 3.48. No surprise there. Trucking treated telematics as safety equipment long before the ELD mandate made it universal in December 2019.

Utilization is where it gets interesting. Every one of ATRI’s top ten “industry KPIs” — high importance and high use — is a safety metric. Critical events, 92.8% of fleets. Hard braking, 92.8%. Speeding, 91.0%. Harsh turning, 90.1%. Harsh turning scored 3.57 on importance, near the bottom of the safety list, and nine fleets in ten track it anyway.

Then look at operations. Cost per mile got the highest importance score in the category, 3.88, and 27% of fleets don’t use it. ATRI’s own line: “Without knowing cost per mile, it is impossible to accurately determine whether a load will be profitable.” Revenue per mile, 71.2%. Loaded versus empty miles, 67.6%. Revenue per pound hauled, 49.5%.

Maintenance is worse. The most-used maintenance KPI, preventive maintenance scheduling, sits at 77.5% — lower than the least-used safety KPI. Missed PM, second-highest importance in its category, is tracked by two-thirds of fleets. Tire and brake health, about half. Downtime prediction, 49.5%.

This is a convenience sample of 111 carriers, mean fleet 1,910 trucks, median 92, skewed to truckload dry van and reefer. Read it as direction, not decimals. The direction isn’t subtle.

Count the systems, not the sensors

ATRI doesn’t frame the table this way, but read it by how many systems each number has to cross before it exists.

KPI	Where the number is born	Importance (of 5)	Fleets using it
Hard braking	Telematics device alone	3.89	92.8%
Speeding	Telematics device alone	4.23	91.0%
Fuel use / MPG	Telematics device (engine data)	3.71	88.3%
Idle / dwell time	Telematics device alone	3.45	86.5%
Preventive maintenance scheduling	Telematics + maintenance software	3.97	77.5%
Cost per mile	Telematics + TMS + accounting	3.88	73.0%
Loaded vs empty miles	Telematics + TMS	3.37	67.6%
Tire / brake health	Telematics + TPMS hardware + maintenance software	3.41	52.3%
Vehicle downtime predictions	Telematics + maintenance history + a model	3.11	49.5%

Importance and utilization: ATRI, September 2026, n=111. The “where the number is born” column is my read of what each KPI needs, not ATRI’s.

Hard braking is one accelerometer and one threshold. The telematics platform computes it, alerts on it, and puts it on a scorecard before anyone at the carrier lifts a finger. Cost per mile needs odometer miles from the device, load and revenue from the TMS, and fuel, payroll, tolls and repair spend from accounting — joined per tractor, per month, with a key that agrees across all three. That join is the product. Nobody sells it, because it depends on how your TMS names a tractor and how your accounting system codes a repair.

The integration numbers in the same report tell you why the join rarely exists. 84% of respondents run a TMS, but 29% run one that isn’t connected to telematics. 48% don’t connect telematics to maintenance software at all. 39% don’t connect it to business transaction data — billing, accounting — and 8% do that connection through spreadsheets. 21% of fleets dispatch by text and phone, which ATRI notes produces “no structured data” and therefore no straightforward way to measure anything downstream of a dispatch.

ATRI’s conclusion puts it carefully: adoption of operational and maintenance KPIs “may be partially dependent on a carrier’s ability to integrate telematics data with business transaction and maintenance tracking systems.” The telematics providers ATRI interviewed for the same report were blunter about the cause — organizational silos between safety and finance, “data overload,” insufficient IT capacity, and an “organizational capacity gap” that needs people who can think like a data analyst and a trucker at the same time.

I’ve built this join before, in a different industry, between a CRM, a field-service platform and an ERP — the seam where the data contract lives. The failure mode is always the same and never glamorous: three systems, three IDs for one physical asset, a timestamp that’s UTC on the device and local on the dock appointment. Same telematics. Different integration.

Detention: measured, billed, unpaid

If you want the cleanest case of a number being captured and never consumed, it’s detention.

Fleets do track it. Idle/dwell time is the second most-used operations KPI at 86.5%. Detention time itself, 73%. And ATRI’s September 2024 detention study found drivers reported being detained on 39.3% of stops in 2023, that for-hire trucking lost more than 135 million hours to it, and that the industry lost $3.6 billion in direct expenses plus $11.5 billion in lost productivity. 94.5% of fleets charge detention fees. They’re paid on fewer than half of those invoices.

So the truck’s geofence knows when it arrived. The TMS knows the appointment time. The rate confirmation says two free hours. And the invoice still goes out late, gets disputed, and goes unpaid more often than not, because the three facts live in three systems and the person assembling the claim is too often doing it from a phone call with the driver a week later.

The measurement itself is shaped by the billing rule, not the operation. The DOT Inspector General’s January 2018 audit of detention found that “accurate industrywide data on driver detention do not currently exist because most industry stakeholders measure only time spent at a shipper or receiver’s facility beyond the limit established in shipping contracts — often 2 hours.” Nobody records the first 119 minutes. The same audit estimated that a 15-minute increase in average dwell time raises the expected crash rate by 6.2%. The number that matters for safety starts at minute one. The number that gets recorded starts when the invoice does.

Meanwhile the automation is arriving at the part of the dock that was already legible. Opendock says its dock-scheduling MCP server has been live since June 2026, letting an AI agent book, move and cancel appointments with no per-warehouse integration. That’s a vendor describing its own product, so weigh it accordingly — but notice what it’s a description of. Booking has a schema. Detention disputes, by Opendock’s own account, still need a human. Agents amplify whatever was already structured and skip whatever wasn’t — the same pattern as when a carrier exposes its automation stack as an API and the mess moves to the prep layer.

What the working version looks like

Not a dashboard. Fleets have dashboards. The telematics provider’s fleet management system — which 78% of respondents use — already has more charts than anyone opens.

The working version is a pipe with an owner. Three pieces.

One asset key. The telematics unit ID, the TMS tractor number and the maintenance system’s asset ID map to one truck, in one table, maintained by one named person when a unit is swapped. It sounds trivial. It’s the thing missing from most multi-system integrations I’ve cleaned up, and without it every KPI in the bottom half of ATRI’s table is a manual reconciliation.

Events written to the record that makes the decision, not to a report about it. Geofence entry and exit timestamps land on the load record in the TMS, next to the appointment time, in the same time zone. Detention becomes a computed field — dwell beyond the free time on that rate confirmation — and the invoice goes out the day the load delivers with the timestamps attached as evidence, not the week after. Engine fault codes land on the asset’s maintenance record as a work-order candidate — converting alerts into work orders automatically is the practice the providers in ATRI’s scan credited with fewer breakdowns.

A monthly cost-per-mile job that nobody has to run. Odometer miles per tractor from telematics, joined to fuel card, driver pay, tolls and repair orders by that one asset key, divided out, compared against ATRI’s 2025 industry average of $2.336 per mile — the highest in the report’s history, up 3.4% in a year — and sorted by lane and customer. That’s the number that decides which freight to stop hauling. It’s also the number 27% of fleets can’t produce.

None of this is AI. It’s the integration and data-contract work that decides whether anything smarter is possible later. A fleet that wants an agent to renegotiate detention terms with a shipper first needs a detention record the agent can read, and most don’t have one.

Hard braking got wired to a decision because it only needed one system, and the platform did the wiring. Cost per mile needed three, and the wiring was the fleet’s job.