How AI Load Matching Works

AI load matching scans load boards continuously, scores every posting against a carrier's own criteria — preferred lanes, minimum rate, equipment type, truck position, deadhead tolerance, broker reliability — and surfaces the highest-value loads in seconds. Instead of a dispatcher scrolling DAT for two to three hours a day and eyeballing a few dozen loads an hour, the system evaluates the whole board and ranks what it finds by profitability, positioning, and backhaul potential. Numeo Spot brings this into the DAT and Truckstop load boards as a Chrome extension, scoring each posting in place so the dispatcher reviews ranked loads instead of raw listings.

The DAT network alone carries more than 500,000 loads posted a day across 1.7 million trucks. A dispatcher scanning that volume by hand is drinking from a fire hose. AI load matching turns the fire hose into a filtered pipeline — the handful of loads a day that actually fit a specific carrier's trucks, lanes, and profitability targets.

Why Manual Load Finding Breaks Down

The core problem is not that loads are hard to find. On a busy day the DAT network alone holds 500,000-plus postings. The problem is that finding the right loads for a specific truck at a specific time and rate takes more time and cross-referencing than manual search can deliver.

The volume problem

A dispatcher managing 20 trucks needs profitable loads for each one, every day. That means weighing dozens of options per truck across origin, destination, rate, pickup time, equipment, broker reliability, and deadhead. At 50 to 100 loads reviewed per focused hour, a dispatcher might cover 400 to 600 loads in a full shift — a sliver of what is posted, and never the same sliver twice.

The loads a dispatcher never sees are the ones that cost the carrier the most. A posting at $2.45 a mile on a preferred lane can sit on the board for half an hour before another carrier books it, missed only because the dispatcher was stuck on a check call.

The speed problem

Spot loads are perishable. A well-priced load on a popular lane can be gone within minutes of posting, and a second phone call often means a second too late. Booking a single load by hand — search, vet, call, negotiate, confirm — can eat a large chunk of a shift across a full board.

AI does not carry that latency. It watches the boards in real time, scores every new posting against the carrier's criteria the instant it appears, and flags the high-value ones before a competitor's dispatcher has finished scrolling.

The data problem

Evaluating a load properly means cross-referencing several data points at once:

• Current market rate for the lane (is the offered rate above or below market?)

• Deadhead miles from the truck's current position to pickup

• Fuel cost for the route (diesel prices vary regionally)

• Toll costs on the specific route

• Broker payment history and reliability

• Load-to-truck ratio on the lane (is capacity tight or loose?)

• Backhaul potential from the destination (will the truck be stranded?)

• Hours of service remaining for the driver

Checking these by hand can run 5 to 15 minutes per load. Multiply that across 20 to 30 candidates per truck and the math eats the whole shift before a single broker call gets made.

How the Algorithms Work

Modern AI load matching follows a two-stage architecture, much like the recommendation systems behind e-commerce: first filter the universe of loads down to plausible matches, then score and rank those matches by value to the carrier.

Stage 1: candidate generation

The system filters the full board down to a workable set of candidates using hard constraints:

• Equipment match: Only dry van loads for a dry van truck, only reefer for reefer, only flatbed for flatbed. Weight and dimension requirements must align.

• Geographic feasibility: The truck must be able to reach the pickup location within the specified window given its current position and the driver's remaining hours of service.

• Minimum rate threshold: Loads below the carrier's floor rate are excluded immediately.

• Broker exclusions: Carriers can blacklist specific brokers based on past payment issues or disputes.

This stage drops the vast majority of postings in milliseconds, cutting hundreds of thousands of loads down to a few hundred relevant candidates per truck.

Stage 2: scoring and ranking

The surviving candidates get scored across several weighted factors. Each load earns a composite profitability score, and the system ranks them from highest to lowest value for that specific truck.

Revenue per mile, loaded plus deadhead. The anchor factor. The system divides total pay — linehaul plus fuel surcharge plus accessorials — by total miles, counting the deadhead from the truck's current position to pickup. A load paying $2.50 a mile with 200 miles of deadhead is worth less than one paying $2.30 with 20.

Deadhead to pickup. The exact distance from the truck's current position to the pickup. Shorter deadhead means higher effective revenue and lower fuel burn. With empty miles running an estimated 15 to 30 percent of total miles industry-wide, and ATRI's 2025 cost research (on 2024 data) putting a truck's marginal cost near $2.26 a mile, every empty mile is real money earning nothing.

Market rate comparison. The system pulls the current lane rate from DAT or Truckstop and measures the offer against it. A load at $2.40 a mile where the market sits at $2.20 scores above an identical $2.40 load where the market is $2.50 — the first is above-market value, the second below.

Lane familiarity. The model tracks the carrier's own booking history. A lane the carrier has run profitably dozens of times scores higher than an unfamiliar one, reflecting the real operational edge of known routes, receivers, and reload markets.

Backhaul positioning. The system does not score a load in isolation. It weighs where the delivery drops the truck. A load into Atlanta, a dense outbound hub, scores above an identical-rate load into a rural market with little freight coming out.

Broker reliability. Scoring folds in payment history, factoring data, days-to-pay, and dispute frequency. A $2.40 load from a broker who pays in 15 days with no disputes beats a $2.50 load from one who pays in 45 and reprices freight after the fact.

Load age. How long a posting has sat signals broker flexibility. A load up for 30 minutes is fresh and the broker is firm; one sitting for six to eight hours suggests they have struggled to cover it and will negotiate.

Time sensitivity. A tight pickup window the truck can just make is an opening most carriers skip — fewer competitors, more leverage.

How the model learns

The scoring is not static. The model refines its weights over time from outcomes:

• Loads the dispatcher accepted vs. rejected teach the algorithm about real preferences beyond stated criteria

• Booked rates vs. initial offers teach the algorithm about negotiation patterns on specific lanes

• Completed loads vs. falloffs teach the algorithm about which broker and lane combinations are reliable

• Revenue per mile outcomes across different scoring thresholds teach the algorithm to calibrate its profitability predictions

After several hundred loads, the model reflects how that particular operation actually makes money, not just generic market averages.

What Load Matching Changes

More of the board, evaluated

The core advantage is coverage. A dispatcher reviewing a few hundred loads a shift misses most of what is posted; the system reads the whole board continuously. The point is not raw volume — it is that the loads a dispatcher never sees are exactly the ones that would have paid best. Surface even a few high-value loads a day that would otherwise have slipped past, book a share of them, and the gain compounds across a month into loads, and revenue, that simply were not on the table before.

Fewer empty miles

Empty miles are the quiet drain on carrier profitability. With deadhead running an estimated 15 to 30 percent of total miles industry-wide and a marginal operating cost near $2.26 a mile (ATRI, 2025, on 2024 data), miles run empty cost nearly as much as loaded ones while earning nothing. Backhaul scoring attacks this directly: by weighting each load on where it leaves the truck, the system steers toward destinations with freight coming back out, so fewer empty legs get booked into the plan in the first place. Even a few points off a carrier's empty-mile percentage, held across a year and a fleet, is real money kept.

Better-paying loads, not just more

Matching does not only surface more loads — it surfaces better ones, by scoring for above-market rates and strong lane positioning. When the system ranks the highest-value loads to the top instead of leaving a dispatcher to book the first decent option they spot, the average booked rate drifts up. A few cents per mile sounds small, but across hundreds of loads a month it is the margin between a thin year and a good one.

How Numeo Does It

Numeo runs load matching from inside the dispatcher's existing workflow rather than asking them to switch platforms.

Numeo Spot is a Chrome extension that layers on top of the DAT and Truckstop load boards. As the dispatcher scrolls, it overlays each posting with the data they would otherwise look up by hand:

• Per-load revenue-per-mile and margin after fuel, tolls, and deadhead

• Broker reliability signals from payment history and factoring data

• Filters that surface loads matching the carrier's criteria

• One-click routing with toll-cost data for an accurate per-load cost

• Drafted broker outreach with load details already filled in

The dispatcher never leaves the board. For carriers ready to automate the next step, the AI Hub takes scored loads further — ranking them against the carrier's rules and drafting the broker negotiation by email — while every booking stays a dispatcher's call.

Manual search vs. AI matching

The comparison is not AI against dispatchers. It is dispatchers with the tool against dispatchers without it: one reviews pre-scored, pre-ranked loads with the profitability already worked out; the other scrolls raw listings and hopes to catch the good ones in time.

Why the Carrier Side Matters

Most AI matching investment has gone to the broker side, and those tools optimize for the broker's goal — finding the lowest-cost carrier who will haul reliably. A carrier's goal is the mirror image: the highest-paying load that fits the truck's position and constraints. An algorithm trained on broker data optimizes for broker margins, not carrier profitability, which is why carrier-first tooling is a different product and not just the same engine pointed the other way.

That asymmetry is the opening. The freight market runs on roughly 787,000 active carriers (FMCSA, December 2023), the vast majority of them small fleets running ten trucks or fewer (ATA, 2025) — exactly the operators who have had the least access to matching software built for their side of the deal. Closing that gap is both the gap and the opportunity.

Getting Started

Install Numeo Spot on your DAT or Truckstop board and let it score loads in place — RPM and margin after fuel, tolls, and deadhead, broker reliability, and routing, all in context as you scroll. Run it for a couple of weeks against your normal process, then compare the numbers that matter: loads booked per day, average revenue per mile, empty-mile percentage, and hours spent searching. If the ranked board is finding loads you were missing, the AI Hub is the next step up — it carries scored loads into drafted, dispatcher-approved broker negotiation.