Digital Freight Matching: How Freight Matching Algorithms Pair Loads and Trucks
Digital freight matching software pairs truckloads with available carrier capacity using data and optimization. This guide explains how the matching works, where it happens, what data it needs, how to measure it, and why some loads still sit unassigned.


Freight matching is the job of finding a truck for a load, or a load for a truck. For most of the history of road freight it was done by phone and email: a shipper or broker called carriers they knew, asked who had a truck near the pickup, and negotiated a rate one load at a time. Load boards moved the listings online but left the matching to the people reading them.
Digital freight matching (DFM) automates that step. Software scores the possible pairings between loads and available trucks, then proposes the best ones to a planner or books them directly. This guide explains how the algorithms work, where matching happens, what data it needs, how to measure it, and why loads still sit unassigned.
§ 01What is digital freight matching?
Digital freight matching is the use of software to pair freight demand with carrier supply. On the demand side are loads, each with an origin, a destination, pickup and delivery windows and an equipment requirement. On the supply side are trucks and drivers, each with a position, an availability time and remaining driving hours. The output is an assignment: which carrier takes which load, and at what price.
Most digital freight matching is applied to full truckload freight, where one load fills one truck and the pairing is one-to-one. The difference from a load board is who does the matching. A load board is a list that carriers search. A matching system ranks the options, often offers a load first to the few carriers most likely to accept it, and records the outcome so that later matches can use it.
§ 02How freight matching algorithms work
At its core, freight matching is an assignment problem. Loads sit on one side and trucks on the other, and every feasible pair gets a score built from the empty distance to the pickup, the chance of arriving inside the pickup window, the price and the carrier's service record. The goal is the set of pairs, each load and truck used at most once, with the best total score. In graph terms this is weighted bipartite matching, which can be solved exactly. The Hungarian method, published by Harold Kuhn in 1955, is the classic algorithm, and modern solvers handle thousands of loads and trucks.
Feasibility comes first. A truck is a candidate for a load only if it has the right equipment, such as a dry van, reefer or flatbed, can reach the pickup in time, and can complete the trip within driving-time rules. In the United States, property-carrying drivers may drive at most 11 hours after 10 consecutive hours off duty, and not beyond the 14th hour after coming on duty. In the EU, daily driving is limited to 9 hours, extendable to 10 hours twice a week, with a break of at least 45 minutes after no more than 4.5 hours of driving. A matching engine that ignores these limits proposes matches that fail in execution.
Deadhead, the distance a truck drives empty, is where most of the value lies. A greedy approach gives each new load to the closest available truck. That minimizes empty driving for this load but can leave the next one without a nearby truck, and it ignores what happens after delivery. Better engines look ahead: they weigh the chance of a backhaul near the delivery point and solve matches in batches over a short time window. When a truck takes a chain of loads, the problem becomes vehicle routing and scheduling, usually solved with mixed-integer optimization or heuristics.
Pricing is part of the match. A rate that is too low leaves a load unaccepted, and a rate that is too high erodes margin. Platforms estimate a market rate for each lane from past and current transactions, then adjust it for lead time, equipment and how hard the load is to cover. Some offer a slightly higher rate on a load that removes an empty leg elsewhere in the network, because the network as a whole saves more than the premium.
Finally, a match only counts if the carrier accepts it. Matching engines therefore estimate the probability of acceptance from each carrier's history: the lanes they run, the rates they have accepted and the days and times they prefer. Offering a load first to the carriers most likely to take it shortens the time to cover and reduces the number of rejected tenders.
§ 03Marketplaces, brokers and TMS: where freight matching happens
Digital freight marketplaces connect many shippers with many carriers on one platform and run the matching centrally. Their advantage is liquidity: with more loads and trucks in the pool, the algorithm has more chances to find pairings with little empty driving. Digital brokers have the same commercial role as traditional brokers, taking responsibility for the load and paying the carrier, but they use matching software to cover loads with fewer manual calls.
A shipper's transport management system (TMS) usually tenders loads to contracted carriers in a set order, known as a routing guide, and falls back to the spot market when they decline. Matching inside a TMS works on a smaller pool, the shipper's own carriers, but with better data on the shipper's own freight. Carriers with their own fleets face the mirror problem of finding loads for their trucks, which is close to fleet dispatch and routing. Many organisations use all three channels, so the practical question is where each matching decision is made and on what data.
§ 04What data freight matching needs
Good matching depends more on data quality than on the algorithm. The minimum set is: load details (origin, destination, pickup and delivery windows, weight, equipment and special requirements such as dangerous goods or temperature control); live truck positions and availability, usually from telematics or electronic logging devices; drivers' remaining hours; carrier profiles covering equipment, operating area, insurance and compliance status; historical rates by lane and date; and the history of tenders, acceptances and rejections. Travel time estimates should reflect real road speeds and traffic, since a match that looks feasible on straight-line distance can miss its pickup window.
The most common gap is stale availability. If a truck shows as free but was booked an hour ago through another channel, the engine recommends a match that will be rejected. Integrations that update truck positions and status automatically are usually worth more than refinements to the scoring model.
§ 05KPIs for digital freight matching
Time to cover: the time from load creation to carrier confirmation, the most direct measure of matching speed.
First-tender acceptance rate: the share of loads accepted by the first carrier offered. A low rate points to pricing or carrier selection.
Empty mile ratio: empty distance as a share of total distance, where better matching shows up in cost and emissions.
Cost per load against a lane benchmark, to check that faster matching is not simply paying more.
On-time pickup and delivery, which shows whether the matches were realistic.
Late-covered and unassigned loads near pickup time. Track all of these by lane and day of the week, because network averages hide the lanes where matching fails.
§ 06Why loads sit unassigned
When loads stay unassigned until the last minute, the cause is usually one of five things. Thin lanes: on low-volume lanes and remote origins there may be no truck nearby at the right time, and no algorithm can create one. Short lead times: loads posted a few hours before pickup leave few options.
Rates below market: a benchmark that lags a tightening market produces offers carriers decline. Hard constraints: unusual equipment, narrow appointment windows or dangerous goods shrink the pool of eligible carriers. Poor data: stale truck status, incomplete carrier profiles or wrong pickup windows lead to recommendations that fail.
The fix depends on the cause. Thin lanes call for combining loads into round trips or releasing them earlier, pricing problems call for faster rate updates, and data problems call for better integrations. Breaking unassigned loads down by lane, lead time and gap to market rate usually shows which cause dominates.
When a shipper or carrier runs its own fleet, the matching decision merges with dispatch and routing. Predictim Routing, our vehicle routing platform and API, computes multi-stop routes across mixed fleets in real time and enforces delivery windows and driver-hour rules as hard constraints, which is the planning layer under fleet-side freight matching.
§ 07What is the difference between digital freight matching and a load board?
A load board lists loads and trucks for people to search and contact. Digital freight matching ranks or selects the pairings automatically, using location, timing, equipment, price and acceptance history.
§ 08Is a digital freight matching platform the same as a freight broker?
A broker is a business that arranges transport and takes responsibility for the load. Freight matching is the decision process of pairing loads with trucks, and brokers, marketplaces, shippers and carriers can all use matching software.
§ 09What algorithm is used for freight matching?
The basic model is weighted bipartite matching, also called the assignment problem, which can be solved exactly with methods such as the Hungarian algorithm. When trucks take chains of loads, the problem becomes a form of vehicle routing and is solved with mixed-integer optimization or heuristics.
§ 10How does digital freight matching reduce empty miles?
It selects trucks close to the pickup and looks for follow-on or backhaul loads near the delivery point, so fewer trucks drive empty between jobs. The effect is strongest on busy lanes where many loads and trucks are available at the same time.
§ 11Does digital freight matching work for LTL freight?
Yes, with a different model. Less-than-truckload freight shares trucks, so matching becomes a consolidation and routing problem with many shipments per vehicle.
Not ready for a project? Ask an engineer your question by email. No call needed.
Ask a question

