How to Reduce Deadhead Miles for Asset Carriers

If you run your own trucks, you already know how to reduce deadhead miles for asset carriers in theory: keep the next loaded mile lined up before the wheels go empty. In practice it does not happen, because the backhaul opportunity lives on a load board while the constraints that decide whether the driver can actually take it, hours of service remaining, current location, equipment type, and dwell ETA, live in a different system entirely. So a dispatcher does the join in their head, under time pressure, across three screens, and the truck rolls empty more often than anyone wants to admit. Empty miles for the industry sat at 16.7 percent in 2024 per ATRI, and plenty of fleets run worse.

Why do asset carriers run 15 to 25 percent deadhead?#

Asset carriers run 15 to 25 percent deadhead because the next load and the constraints that qualify it live in different systems, and no one joins them while the join still matters. The load board knows what freight is available. It does not know that the driver heading toward Laredo has four hours left on the 11-hour driving clock, is pulling a reefer when the available load needs a dry van, or will not actually clear the receiver until 3 p.m. tomorrow. Those facts live in the ELD and the TMS. The match that would keep the truck loaded requires all three data sources at once, and they are never in one place when the dispatcher needs them.

The cost of getting this wrong is not abstract. ATRI's 2024 operational cost research put the average marginal cost of operating a truck at $1.779 per mile excluding fuel, the highest non-fuel figure ATRI has ever recorded, with empty miles averaging 16.7 percent across the industry. Every one of those empty miles costs roughly the same to run as a loaded one, minus the fuel, and earns nothing. FreightWaves frames the benchmark plainly in its deadhead reduction coverage: under 15 percent is sharp, and consistently over 20 percent means the freight mix is broken. A fleet sitting at 22 percent empty is not lazy. It is flying blind on the one decision that determines utilization.

The structural cause is that the backhaul decision is a join across systems that were each bought to do one job well. The load board does freight discovery. The TMS does the commercial load and the appointment. The ELD does the truck and the hours. Each is right about its own slice. None of them sees the other two, so the person who has to see all three is a dispatcher with a phone, a spreadsheet, and twenty trucks to keep loaded before end of day. That person makes good decisions when they have time and bad ones when they do not, and at scale they never have time.

What data do you actually need to match a backhaul to a driver?#

You need four things joined at the same instant: the driver's current location, the hours of service remaining, the equipment type on the trailer, and the realistic ETA out of the current stop. A backhaul match that ignores any one of these is a guess. The available load on the board might be a perfect lane and still be unworkable, because the driver cannot legally reach it, cannot haul it with the trailer they have, or will not be free in time to make the pickup window.

Hours of service is the constraint that quietly invalidates most matches, and it is the one a load board cannot see. Under the FMCSA hours-of-service rules in 49 CFR 395, a property-carrying driver gets 11 hours of driving inside a 14-hour on-duty window after 10 consecutive hours off. A backhaul that looks ideal on a map is illegal if it puts the driver past the 14-hour clock. That information lives in the ELD, updated continuously as the driver works. It is the single most decision-relevant fact for a backhaul match, and it is structurally absent from the place where backhauls get discovered.

Equipment type is the second silent filter. An asset carrier's fleet is not interchangeable. A reefer cannot take a dry van load, a flatbed cannot take a van load, and a 53-foot trailer cannot serve a dock built for a 48. The equipment assignment lives in the TMS, tied to the unit and the driver. The dwell ETA, when the truck will actually be free, depends on the current stop's real progress, which lives in the appointment record and the telematics gate timestamps. Werner, Schneider, and J.B. Hunt all run mixed fleets across reefer, dry van, and intermodal precisely because different freight needs different equipment, which means equipment matching is not an edge case for an asset carrier. It is the baseline.

What is ACT mode, and how does a backhaul agent use it?#

ACT mode is the intelligence mode where an agent closes a routine cycle end to end and a human stays on the exceptions. In Mithrilis terms it answers "handle it for me." A backhaul agent in ACT mode does the join a dispatcher would do by hand: it reads the driver's live location and hours from the ELD, the equipment and the dwell ETA from the TMS, and the live load opportunities from the board, and it surfaces only the loads that legally and physically fit. The dispatcher does not search. The dispatcher approves.

This is the distinction that matters. ACT mode does not mean the agent books loads while the dispatcher sleeps. It means the agent does the part humans do badly, the continuous reconciliation of four data sources across twenty trucks, and the human does the part humans do well, the judgment call on a borderline match or a new lane. The agent narrows a thousand available loads to the three that fit this driver, in this location, with these hours, on this trailer, by this ETA. A person who used to spend the afternoon hunting now spends it deciding. That is the trade ACT mode makes: the routine cycle closes itself, and human attention moves to the exceptions where it is worth something.

It also means the intelligence comes from connected data, not from automating one workflow inside one tool. A load board with a better filter is still a load board. It cannot see the HOS clock. The agent is valuable precisely because it sits across the systems and joins what none of them can join alone. This is the Mithrilis thesis applied to deadhead: connect the systems you already run, resolve the picture, and let intelligence act on the resolved picture with a human in the loop. The same connect-and-resolve foundation is described in our guide on unifying TMS, ELD, and dock data without replacing your TMS, and the full approach lives on the Mithrilis platform.

How does a backhaul agent match HOS and equipment to a live load?#

A backhaul agent matches by treating the driver's constraints as hard filters on the live load set, then ranking what survives. Walk through one truck. A Schneider-style reefer unit delivers in Laredo, Texas at 11 a.m. and is scheduled to clear the receiver, per the appointment record and the live gate timestamps, by 1 p.m. The driver has 6 hours and 40 minutes left on the 11-hour driving clock and is well inside the 14-hour on-duty window. Without a backhaul, that truck deadheads back toward the home domicile in San Antonio and then keeps running empty toward the next assignment, burning paid miles that earn nothing.

The agent runs the join the moment the dwell ETA firms up. It takes the Laredo location, the 6 hours 40 minutes of remaining drive time, and the reefer equipment type, and it filters the live load board to loads that originate within legal driving range of Laredo, need a reefer, and have a pickup window the driver can make before the clock forces a 10-hour break. A produce load from the Rio Grande Valley moving north toward Dallas survives all three filters. It is reefer freight, it picks up inside the driving window, and it points the truck in a direction that sets up the next loaded mile instead of a dead one. The agent surfaces it, with the HOS math and the equipment match shown, and the dispatcher approves the booking in one click.

What makes this trustworthy is that every input is visible. The proposed match shows the ELD clock it read, the TMS appointment it used for the dwell ETA, and the load record it pulled from the board. If a dispatcher distrusts the hours figure, they click through to the ELD source. If the equipment looks wrong, they see which TMS field the agent read. Sources are visible and the match is inspectable, which is what lets a human approve at speed instead of re-checking the agent's work from scratch. A match you cannot verify is a match you will not trust, and a dispatcher who does not trust the agent will go back to the load board and the phone, which is exactly the manual process the agent was supposed to retire.

How does the same backhaul intelligence help a freight broker?#

The same backhaul intelligence lets a freight broker reposition a carrier's empty capacity into a paying load instead of letting it run home dry. A broker does not own the truck, but a broker who knows a contracted carrier's truck is about to deadhead out of Laredo with hours to spare holds a real opportunity: cover a load that needs covering with capacity that is already moving, at a rate that beats what the carrier would otherwise earn on the empty leg, which is nothing. The constraint is identical to the carrier's. The broker still needs the driver's location, the hours of service, the equipment, and the realistic ETA before offering the load, or the offer is a guess that wastes everyone's time.

The difference is whose systems hold the data and how it gets shared. For an asset carrier the four facts live in its own ELD and TMS. For a broker the same facts arrive through the carrier relationship and the visibility feeds the broker already consumes. In both cases the intelligence is the join, and in both cases the agent surfaces the workable matches and a human approves the booking. A broker running this against a roster of contracted carriers turns a fleet of soon-to-be-empty trucks into a live capacity pool, matched lane by lane to freight that needs covering. That is the same ACT-mode cycle, pointed at a different owner of the asset. The mechanics for each audience are detailed on the freight brokers and asset carriers pages.

This is why the deadhead problem is not broker-versus-carrier. It is a shared structural gap. The carrier wants its own trucks loaded on the way home. The broker wants to monetize capacity that is already in motion. Both are blocked by the same thing, the failure to join the load opportunity to the live constraints, and both are unblocked by the same intelligence. The market sets these two up as opposites, but on the deadhead question their interests point the same way: keep the truck loaded, and split the value of the mile that would otherwise have run empty.

What does reducing deadhead actually change for an asset carrier?#

Reducing deadhead changes the economics of a fleet you already own without adding a single truck. The asset is fixed. The driver is paid whether the truck is loaded or empty. The payment, insurance, and depreciation accrue per mile regardless. So every empty mile converted to a loaded mile is close to pure contribution, which is why deadhead is the highest-leverage operational number an asset carrier has. ATA's American Trucking Trends 2025 reported that trucks moved 11.27 billion tons of freight in 2024 across an industry where 91.5 percent of carriers run ten or fewer trucks, which means the typical asset carrier is a small fleet for which a few points of deadhead reduction is the difference between a profitable quarter and a flat one.

Consider the math on a single truck. A unit running 100,000 paid miles a year at 22 percent deadhead is running 22,000 empty miles. Pull that to 14 percent and 8,000 of those miles become loaded, each one carrying the marginal cost the truck was already going to incur. At the ATRI marginal cost of roughly $1.78 per mile excluding fuel, those recovered miles stop being a pure loss and start carrying revenue. Multiply across a fleet and the deadhead number stops being a metric on a report and becomes the lever that moves the operating margin, which ATRI found sat below 2 percent in every truckload sector in 2024. The fleet did not get bigger. The same trucks just ran loaded more often.

The change is also a change in where dispatcher attention goes. Today a dispatcher spends the afternoon hunting the board and reconciling it against the TMS and the ELD in their head. With an ACT-mode agent doing that reconciliation continuously, the dispatcher's day shifts from search to judgment. They handle the exceptions, the borderline HOS calls, the new shipper relationship, the load that pays well enough to justify a slightly longer reposition, and the agent handles the routine matches that used to eat the day. The Bureau of Transportation Statistics tracks national freight movement through its Freight Facts and Figures program; the macro picture it describes is the sum of millions of these individual load-or-empty decisions, made better or worse one truck at a time. Getting the per-truck decision right is what the agent is for, and it is what closes the gap between the deadhead a fleet runs and the deadhead it could.

None of this requires replacing the TMS, the ELD, or the load board. The agent reads from the systems already in place, joins what they cannot join alone, and acts on the resolved picture with a dispatcher approving every booking. That is intelligence from connected data, which is the whole point of how we think about the product: the value was never in any one system, it was in the join nobody had time to make.

What it takes to actually put empty miles to work#

Reducing deadhead is not a motivation problem. Every carrier wants the next loaded mile lined up. It is a join problem: the opportunity lives on a load board while the constraints that decide whether a driver can take it live in your TMS, ELD, and driver app. Asking a dispatcher to do that join in their head, under pressure, is why the truck rolls empty.

Mithrilis does the join. It connects the systems a carrier runs into one operating picture where hours of service, location, equipment, and dwell ETA sit beside live backhaul options. An agent surfaces the legal, profitable match the moment a load frees up, and Atlas answers which trucks and lanes are bleeding empty miles this week.

The result is fewer empty miles without another screen for your planners to babysit. The intelligence does the repositioning math so your people make the call. See it on your fleet.

Related Mithrilis capabilities

• The Mithrilis platform

  How connected data becomes verifiable intelligence that acts.

• For asset carriers

  Backhaul matching across ELD hours, equipment, and live loads to cut deadhead.

• For freight brokers

  Reposition contracted carrier capacity into paying loads before it runs empty.

Frequently asked questions

How do you reduce deadhead miles for asset carriers?

You match each driver's real constraints, current location, hours of service remaining, equipment type, and realistic ETA out of the current stop, against live load opportunities, instead of asking a dispatcher to reconcile a load board with the TMS and ELD by hand. The empty mile is a data-fragmentation problem: the backhaul opportunity and the constraints that qualify it live in separate systems. An ACT-mode agent joins them, surfaces only loads that legally and physically fit, and a human approves the booking.

What is a good deadhead percentage for a trucking fleet?

FreightWaves describes under 15 percent as running sharp and consistently over 20 percent as a sign the freight mix is broken. ATRI put industry-wide empty miles at 16.7 percent in 2024. Asset carriers commonly run between 15 and 25 percent, so most fleets have real room to improve, and the trucks to do it are already on the road.

How does trucking backhaul optimization actually work?

Backhaul optimization works by treating the driver's hours of service, location, equipment, and dwell ETA as hard filters on the live load set, then ranking the loads that survive. A load that fits the lane on a map is still unworkable if the driver is out of hours under FMCSA 49 CFR 395, is pulling the wrong trailer, or cannot make the pickup window. Optimization means joining all four constraints to the available freight before offering a match, which is the join a load board cannot do alone.

Why is empty mile reduction so hard for carriers?

Carrier empty mile reduction is hard because the data needed to make the decision lives in three systems that never get joined in time. The load board knows what freight is available, the ELD knows the driver's location and remaining hours, and the TMS knows the equipment and the appointment. A dispatcher has to hold all three in their head across many trucks under time pressure, and at scale they cannot, so trucks run empty by default rather than by choice.

How does ACT mode differ from just automating dispatch?

ACT mode means an agent closes the routine cycle, the continuous join of location, hours, equipment, and ETA against live loads, while a human stays on the exceptions and approves every booking. It is not unattended automation that books loads on its own. The agent does the reconciliation humans do badly across many trucks, and the dispatcher does the judgment humans do well, the borderline HOS call or the new lane. The cycle closes itself, but a person stays on the decisions that matter.

Does this help with asset utilization for trucking beyond deadhead?

Yes. Asset utilization for trucking is governed by the same join. Once the agent connects the ELD, the TMS, and the load picture, the resolved data also exposes which lanes consistently strand trucks empty, which equipment types sit idle, and which domiciles generate the most dead miles. Deadhead reduction is the first cycle the agent closes, but the connected picture surfaces the structural patterns that determine utilization across the whole fleet.

Can a freight broker use the same backhaul intelligence?

Yes. A broker who knows a contracted carrier's truck is about to deadhead with hours to spare can offer a load that covers freight needing coverage at a rate that beats the empty leg, which earns nothing. The broker needs the same four facts, location, hours, equipment, and ETA, arriving through the carrier relationship and existing visibility feeds. The agent surfaces the workable matches and a human approves, turning a roster of soon-to-be-empty trucks into a live capacity pool.

Do I have to replace my TMS or ELD to do this?

No. The agent reads from the TMS, the ELD, and the load board you already run through their existing interfaces, joins what none of them can join alone, and acts on the resolved picture with a dispatcher approving every booking. Nothing is migrated and no system is replaced. The value was never inside any one tool. It was in the join nobody had time to make.