Freight audit and pay does something genuinely useful: it takes carrier invoices, checks them against your contracted rates, catches the ones that don’t match, and pays the rest.
For most shippers that’s a clear win over approving invoices by eye. The error rates in LTL billing are high enough that the service pays for itself.
But the model has a boundary, and it’s worth being precise about where it sits. Everything an FPA process knows, it learns from a document that arrives after the freight is delivered.
What the invoice can tell you
Quite a lot, and this shouldn’t be undersold.
Did the carrier apply the discount in your agreement? Is the fuel surcharge calculated off the right base and the right index week? Are there duplicate bills against one pro number? Did an accessorial get charged that isn’t in your tariff? Is the minimum charge applied correctly? Did they bill a class other than the one on the bill of lading?
Those are all answerable from the invoice plus the contract, and they’re all worth catching. That’s the four-way match most audit processes run, and it works.
What the invoice cannot tell you
The invoice describes what the carrier says happened. It has no record of what you intended, and no record of what actually happened on the dock.
Whether the class was right in the first place. The invoice shows the class the carrier billed. If it matches your BOL, the audit passes. If both were wrong — because the class was inherited from a shipper who set it years ago and never re-verified it — the audit passes and you overpay on every shipment of that commodity, forever. Nothing in a document comparison catches a shared assumption.
Whether the service you bought was the service you needed. A guaranteed shipment billed as guaranteed passes audit. Whether it needed to be guaranteed is not an invoice question.
Whether the reweigh was correct. By the time an invoice is audited, the pallet is gone. The audit can confirm the carrier billed the reweighed figure consistently with their tariff. It cannot tell you whether the dimensioner captured two shipments as one, because nothing about that is on the document.
Whether the accessorial was actually performed. A liftgate charge that exists in your tariff and is applied at the contracted rate passes every check. Whether a liftgate was used is a fact about a delivery, not a fact about a document.
That last category is the interesting one, because those charges are correct in every way an audit can measure and still wrong.
The timing problem underneath all of it
Each of the above shares a cause: by the time the invoice exists, the freight is delivered, the pallet is broken down, and the people who handled it have moved on to this week’s work.
An audit finding at that point becomes a dispute over a document that both sides have to reconstruct from memory and machine records. You can win those — with a scale ticket, a photograph, a contract clause — but you’re arguing about something neither party can physically examine.
The same finding while the load is in transit is a different conversation entirely. The freight is on a dock. It can be re-measured. The terminal that captured it is reachable and the shift that handled it is still on. The carrier can verify or correct, and the correction happens before an invoice is ever produced.
Same discrepancy. Radically different odds.
This is not an argument against FPA
Worth being clear, because the category has enough noise in it already.
Invoice audit catches things nothing else catches — systematic discount misapplication across thousands of bills, fuel index errors, duplicates. That’s pattern work across a large document set, and it’s exactly what a good FPA process is built for. If you don’t have one, get one.
The point is narrower: an audit that begins at the invoice inherits the invoice’s blind spots. Adding people or software to the back end doesn’t move that boundary, because the boundary is the document.
What “before the invoice” actually means operationally
It means the source of truth is what was quoted and what was on the bill of lading, not what the carrier billed — and that deviations from it get flagged as they occur rather than discovered later.
A reweigh comes back from the carrier’s API while the load is in the network, and it doesn’t match the BOL. That’s a discrepancy right then, not in six weeks. An accessorial appears on a status update that nobody quoted. Same thing. A service level on the tender doesn’t match what the customer asked for. Same thing.
None of that requires waiting for a bill. All of it is knowable while the freight is still moving, and all of it is easier to resolve then.
Where settlement fits
To be explicit, because it’s a fair question: catching discrepancies in transit doesn’t replace paying carriers. Somebody still reconciles and settles.
The difference is what arrives at settlement. If the corrections happened in transit, the invoice that shows up is already right, and settlement is a payment rather than an investigation. That’s the goal — not to audit harder at the end, but to have less to audit.
The short version: freight audit and pay is real and worth having, and it can only see what the invoice shows. The largest recoverable errors in LTL are the ones where the invoice is internally consistent and still wrong.
That’s the half we built Live Audit for — checking the load while it’s still moving, so the correction lands before billing. The rating side of the same engine is here.
So the question worth sitting with: of the discrepancies your audit found last year, how many could still have been physically verified when you found them?