Buyer's Guide

Freight Audit Software: Buyer's Guide for Mid-Market Shippers

Published May 29, 2026

Freight audit software covers the mechanical errors well — fuel surcharge math, dim weight, accessorial stacking, duplicate billing. What it does not cover well: contract-clause interpretation, classification disputes that need physical evidence, and the gray-area accessorials that need a human auditor. The right buyer is a shipper with $5–25 million in annual freight spend and an internal team to operate the platform. If that is not you, software alone will leave money on the table. Your first audit with Eller is free.

What freight audit software does well

Strip away the marketing and freight audit software is, at its core, an arithmetic engine. It is excellent at arithmetic. On a typical mid-market account, a well-configured platform will flag roughly 80 percent of the mechanical billing errors within the first 60 days of clean data. Specifically:

• Fuel surcharge math. Mechanical: pull the Energy Information Administration (EIA) index for the billing week, apply the carrier's published surcharge table, compare to the invoice line. Match or mismatch.
• Dim weight calculation. Mechanical: multiply length by width by height, divide by the dim divisor in the Master Rate Agreement (MRA), compare to billed weight. Match or mismatch.
• Accessorial stacking checks. Mechanical: was the same accessorial billed twice on one shipment? Were a residential surcharge and a delivery-area surcharge both applied where the MRA forbids it?
• Service-guarantee tracking. Mechanical: was the commit time met? If not, the refund is mechanical too.
• Discount tier verification. Mechanical: did monthly or quarterly volume hit the contracted threshold? If yes, was the discount applied?
• Duplicate invoice detection. Mechanical: same Progressive Rotating Order (PRO) number, same shipment date, billed twice.
• Multi-currency normalization. Mechanical: convert international invoices at the agreed conversion date and rate.

These are arithmetic problems. Software wins. If a vendor cannot show you these basics in a 30-minute demo, walk away. The category has matured to the point where the mechanical recovery layer is effectively table stakes — every serious platform does it, and the differences across vendors at this layer are minor enough that they should not drive your purchase decision. What should drive it is what the platform does when the invoice is not a clean arithmetic problem.

What freight audit software does not do well

After working with most of the major freight audit platforms over the last decade, the pattern is consistent: software catches 60 to 80 percent of recoverable money on a clean account. The remaining 20 to 40 percent requires someone who reads contracts.

• Reclassification disputes. A pallet was billed as Class 125 when the shipper claims Class 70. Resolving the dispute requires physical evidence — a reweigh ticket, a photo of the freight, a National Motor Freight Classification (NMFC) commodity code lookup. Software can flag the gap. It cannot win the dispute.
• Accessorial authorization disputes. A "limited access" fee was billed; the bill of lading (BOL) says nothing about limited access. Was the location actually limited access? Was the fee pre-authorized? Software cannot read the BOL the way an auditor does.
• Contract-clause interpretation. When the MRA says "applicable fuel surcharge per carrier table" without specifying which table version, software guesses. A human auditor calls the carrier rep and asks.
• Detention and demurrage disputes. Was the trailer actually held? For how long? Was the delay shipper-caused or carrier-caused? Resolving requires yard logs, driver app Global Positioning System (GPS) traces, and contract-clause interpretation. Software does not have the yard log.
• Gray-area handling. Limited-access fees, peak-season exemptions, General Service Reduction (GSR) exception lists. The MRA might allow exceptions. Software does not know which exceptions you actually negotiated verbally.
• Reading the MRA for negotiation opportunities. Software audits against the contract you have. It does not tell you the contract should be re-cut.

These are interpretation problems. Software loses without human review.

Worked example: $50M annual program, $450K/yr software-only blind spot

A national distributor with a $50M annual freight + parcel program. Two parallel audit approaches benchmarked on the same 12-month invoice population:

The $450K/year delta breaks down to: reclassification disputes needing physical evidence ($165K), accessorial authorization disputes needing BOL judgment ($120K), contract-clause interpretation on ambiguous "applicable surcharge" language ($95K), detention disputes needing yard-log correlation ($70K).

For Fortune 500 enterprise programs at $300M+ annually, this same software-only blind spot compounds to $2.5M–$4M per year of interpretive recovery that automated platforms cannot capture alone. The buyer decision isn't "software or service" — it's whether the human review layer comes from your internal team or a partner audit firm.

Comparison: leading freight audit software approaches

The market is roughly four categories. Each is built for a different buyer. Pricing varies by an order of magnitude across them.

When to buy software vs hire an audit firm

Neither path is universally correct. The decision turns on three variables: annual spend, internal headcount, and appetite for fixed versus variable cost. Most shippers under $10 million in annual freight spend cannot justify the cost or the FTE needed to operate a serious platform. Most shippers above $50 million can. The $10M–$50M middle is where the question is genuinely open, and where a hybrid arrangement — software for high-volume mechanical, firm for gray-area — often wins.

What to ask in a software demo

Demos are scripted. To get past the script, ask these five questions. Listen for hesitation on questions two and four — that is where the gap usually lives.

1. "What is your data ingestion model — Application Programming Interface (API), Electronic Data Interchange (EDI), or Portable Document Format (PDF) parsing?"
2. "How do you handle MRA clauses that are ambiguous or vague?"
3. "What is your flag-to-recovery conversion rate on typical mid-market accounts?"
4. "Can your platform read contract documents, or only invoice data?"
5. "What is the human-in-the-loop workflow when a flag needs judgment?"

What we cannot tell from a software demo alone

Vendor demos are scripted on clean data. The real signal is your first month of your data flowing through the platform. Most software vendors offer 30-day pilots. Use them. Push your messiest invoices through — the ones with handwritten BOL annotations, the ones from your most chaotic carrier, the ones for international lanes. A platform that handles those cleanly is worth a second look. A platform that hides them in an "unmatched" queue is not. Two more pilot tests worth running: feed in a known-bad invoice you have already disputed and resolved, and see whether the platform catches what you caught. Then feed in an invoice you suspect is wrong but cannot prove, and see whether the platform gives you anything actionable or just adds it to the flag queue and waits for your auditor to figure it out.

How Eller Audit handles this

We use software (and AI) for the mechanical heavy lifting — the fuel-surcharge math, dim recalculations, duplicate scans, and service-guarantee tracking. We use human review for contract-clause interpretation, gray-area accessorials, and reclassification disputes. Best of both. Performance-based pricing — software cost is included in the recovery share. You only pay when we recover money. For most mid-market shippers, this structure removes the hardest part of the buy-vs-build decision: there is no platform subscription to defend at budget time, no implementation timeline, no FTE to hire and train. The first audit is free, recovery share kicks in only when money comes back, and contract-clause review is part of the work, not an add-on.

Frequently asked questions

Related pages