Freight Audit and Payment Software: Why Line Level Rating Beats a Spot Check

Why freight invoices are wrong in ways a spot check cannot find

A logistics finance analyst pulls 30 invoices from last month's 14,000 and checks them against the rate sheet. Twenty eight are fine. Two have an accessorial she does not recognise. She queries them, one gets credited, and the sample is declared healthy. Everyone moves on.

The problem with that ritual is that freight billing errors are not randomly distributed, so a random sample is close to useless at finding them. They cluster. One carrier applies a fuel surcharge against the wrong index week for an entire lane. One terminal reweighs and reclasses a product every time because your packaging description is ambiguous. One account gets an address correction fee on every delivery to the same site. A duplicate arrives under a different pro number after a carrier reissues. Each is systematic, repeats hundreds of times, and is invisible to a thirty invoice sample because each instance looks plausible.

The only method that finds them is rating every line. Not checking, rating: independently computing what the invoice should say from the shipment facts and your contract, then comparing. That is a computation problem at a scale that forces real engineering, which is why this category exists as a service industry at all.

Problem 1: your contract is not a rate table

People describe freight contracts as rates. They are not. A less than truckload agreement is typically a base tariff, a discount varying by lane band, freight all kinds classes or an exception list, a minimum charge and an absolute minimum charge that override it, a fuel surcharge indexed to a published diesel price with a lag rule, and an accessorial schedule with caps. Truckload adds mileage basis disputes, since two mileage engines rarely agree, plus stop charges and detention free time. Parcel adds dimensional divisors, zone tables, earned discount tiers that reset, and surcharges that change annually.

Cass Information Systems, Trax Technologies, nVision Global, A3 Freight Payment and enVista all do this work at scale and do it competently. The honest limitation of the outsourced model is that they audit against the rate data you gave them. If a contract amendment did not reach them, or reached them as a PDF that got keyed with a wrong effective date, the audit passes cleanly and you never learn. You also cannot easily interrogate what was not caught, because the analytics available to you are the analytics they chose to expose, and the underlying data model is theirs.

What a custom build does: make the contract a versioned, testable object in your own system, with effective dates, and put a test suite around it. That last part is the piece that never happens in a service arrangement. You take fifty historical shipments whose correct charge you know, and every time a contract is loaded or amended, they run. If a new tariff load breaks a known case, you find out on the day the contract changed rather than nine months into overpaying.

Problem 2: the invoice is checked against nothing but itself

An accessorial charge is a claim about something that happened in the physical world. Detention means a driver waited. A liftgate means one was needed and used. A redelivery means a first attempt failed. A reconsignment means you changed the destination. Auditing those against a rate sheet only confirms the price of the accessorial, not whether it was earned. That is the larger pool of money and it is untouched by rate audit alone.

Providers cannot check this for you because they do not have your operational data. They have the invoice and the rate. Your yard management system knows when the driver checked in and out. Your warehouse system knows the load was not ready. Your order system knows the delivery address was residential from the start, so an address correction fee is questionable. That evidence exists inside your four walls and never meets the invoice.

What a custom build does: join the invoice line to your own operational record before approving it. Detention billed for 3.5 hours against a gate log showing 1.2 hours is an automatic dispute with evidence attached, not a phone call. In our delivery experience this is consistently where the surprising recoveries are, because rate errors have usually been squeezed already and accessorial validation has not been attempted at all. It is also the piece that keeps working, since accessorials are where carriers naturally expand when rates are tight.

Problem 3: coding and allocation is done by people who are guessing

Every invoice has to land somewhere in the general ledger, and in most companies the coding rules are more complex than anyone admits: by business unit, by plant, by product line, by inbound versus outbound, by whether the freight is prepaid or collect, sometimes split across cost centres on a single truck carrying three divisions' goods. Accounts payable clerks apply these rules from a document that is out of date, and errors here do not show up as overpayment. They show up as a business unit manager arguing that his freight cost is wrong, every month, forever.

Providers will code to rules you define, and they do it reliably. What they cannot do is apply a rule that depends on data they do not hold, such as which purchase order line an inbound shipment relates to, or which work order consumed the material.

What a custom build does: derive coding from the shipment's own links rather than from a lookup on carrier or lane. Once the invoice line is joined to your shipment and the shipment to the order, allocation across cost centres is arithmetic instead of judgement, and multi-way splits stop being a manual exception. Accrual gets fixed as a side effect, because you can accrue on shipment rather than on invoice receipt.

Problem 4: disputes die of friction

A dispute that takes an analyst twenty minutes to raise will not be raised for a $60 error, which is exactly why $60 errors are worth repeating. Carriers know the economics of your dispute process better than you do. Meanwhile the disputes you do raise sit in email, with no state, no ageing and no visibility, until someone remembers to chase, and the credit either arrives or does not.

What a custom build does: generate the dispute automatically from the exception, with the computed correct amount, the contract clause relied on and the operational evidence attached, then track it as a receivable with an age. The threshold for disputing drops to near zero because the cost of raising one is near zero. Then measure carrier accuracy by carrier and by error type, and take that measurement into the next negotiation. A carrier who cannot bill correctly is more expensive than their rate suggests, and until you can quantify that in a meeting, nobody will act on it.

Problem 5: you cannot answer the questions the business asks

What did we spend moving product to the northeast last quarter, split by mode, excluding accessorials, compared to the same quarter last year, and how much of the increase was fuel. That question is asked constantly and answered badly, because the data lives in a provider's reporting tool with dimensions that do not match your business.

What a custom build does: keep the rated, coded, disputed and paid detail in your own warehouse with your own dimensions. That is often the strategic reason a large shipper builds. Not the audit recovery, which is measurable and worthwhile, but ending the position where your third largest cost line is understood only through a vendor's dashboard.

What this costs and how long it takes

Across the 2,000-plus projects Digital Heroes has delivered, this category prices as follows. A focused first release, meaning invoice ingestion across your main formats, a rating engine for your two largest modes, contract versioning with a regression test suite, and line level exception queues, runs $100,000 to $200,000 and ships in 16 to 20 weeks. A full platform adding accessorial validation against operational data, general ledger coding and allocation, accrual, payment file generation, carrier dispute workflow with ageing, and a spend analytics layer runs $250,000 to $600,000 phased over 9 to 15 months.

What pushes the number up: modes, since parcel, less than truckload, truckload, intermodal, ocean and air each need their own rating logic. Parcel specifically, because the surcharge landscape changes annually and dimensional rules are unforgiving. Carrier count, as every invoice format differs. Whether you execute payment, which brings banking integration and a higher control bar. And the state of your contract archive, since amendments living in three inboxes is discovery work before any code is written.

What holds it down: keep your provider for payment execution and build the audit and analytics layer first. That split is underrated. You get the interrogable rating engine and your own data without taking on money movement in phase one.

Build versus buy, and the hybrid nobody discusses

Buy if your freight spend is under roughly $10M, or if it is concentrated with a handful of carriers on simple terms. Cass and its peers are efficient at that profile and your recovery would not fund an engineering team. Buy also if you have no operational data to validate accessorials against, because then you are only doing rate audit and a provider does rate audit well.

Build when two or more of these are true. Freight spend is over roughly $40M across several modes. Accessorials are a growing share of your invoices and nobody validates them against what actually happened. Your general ledger coding requires data only your systems hold. You cannot answer basic spend questions without asking a vendor. Or you have already recovered the obvious rate errors and your recovery rate has flattened, which usually means you have exhausted what rate audit can find.

The hybrid is the answer more often than either extreme: build the rating engine and the evidence based accessorial validation, keep an established provider for payment execution and for carriers where volume does not justify your own integration. You own the intelligence, they run the plumbing. Nobody sells this arrangement because it is not in either party's commercial interest to propose it, which is exactly why it is worth considering.

How to choose a developer for freight audit work

Ask them how they would test a rate contract. If the answer does not involve a suite of historical shipments with known correct charges that runs on every contract change, they will build something that produces confident numbers nobody can verify. That regression suite is the single most valuable artefact of the project.

Ask how they handle a fuel surcharge. It is a good probe because it requires an indexed value, a lag rule and an effective date, and a developer who has not built one will treat it as a percentage field.

Ask what they will do about mileage disagreements on truckload. Two mileage engines produce different results for the same origin and destination pair and the contract usually specifies which governs. If they have not encountered this, they have not rated truckload.

Ask what they have integrated. Electronic invoice interchange from carriers, transportation management systems, yard and warehouse systems for accessorial evidence, and your enterprise resource planning (ERP) system for coding and payment are four different problems. Make them name the specific systems in your estate and the specific document types.

Ask about controls if payment is in scope. Approval limits, segregation of duties, duplicate detection across reissued invoice numbers and an audit trail on every amount change are not features to add later. They are the design. And ask who owns the code: at Digital Heroes the client owns the repository from the first commit, and with a system that computes what you pay, any other arrangement is a governance problem as much as a commercial one.