Why Freehand Audits Every Invoice Instead of a Sample

Most freight audits check 15-30% of invoices, not because that's the right amount to check, but because checking every invoice by hand cost more than the errors it caught.

That math has changed. Full coverage no longer requires adding headcount, which means the sample was never really a strategy. It was the limit of what manual labor could afford to check.

‍

Key Takeaways

• A sample-based freight audit checks a fixed percentage of invoices, typically 15-30%, and approves the rest without a line-item review.
• Sampling exists because the fully loaded cost to process one invoice runs $10.89 to $13.11, almost entirely labor. Checking every invoice by hand meant paying that cost on every single line, whether or not it turned up an error.
• One Fortune 10 consumer technology company audited only 33% of its freight invoices for exactly this reason. $263M in accessorial charges passed through unvalidated annually, not because those charges didn't matter, but because auditing all of them at the old cost structure wasn't affordable.
• Full-coverage AI-native auditing recovers 1.5 to 2.5% of total freight spend, money a sampled audit structurally can't reach, since it never checked the invoice in the first place.
• The constraint wasn't judgment. It was headcount. Once checking every invoice doesn't require adding a person for every additional invoice, the reason to sample disappears.

‍

What does a sample-based freight audit actually check?

A sample-based audit reviews a fixed percentage of invoices, commonly 15-30%, and approves everything outside that sample without a line-item check.

The invoices that get picked are usually chosen by a rule, above a dollar threshold, a random draw, a specific carrier, not because they're more likely to be wrong.

That's the part that gets lost in how sampling gets talked about. It sounds like a targeting method. In practice, it's a cutoff.

For how this compares across every audit model, in-house, BPO, and AI-native, see our guide to the four models of freight audit.

‍

{{blue-cta}}

‍

Why did freight audits start sampling in the first place?

Because checking every invoice by hand meant paying full labor cost on every single line, whether or not it contained an error.

The fully loaded cost to process one invoice runs $10.89 to $13.11, almost entirely headcount. Auditing 100% by hand scaled headcount linearly with volume.

A U.S. AP clerk's base salary alone runs $38,500 to $53,500, 1.25x to 1.4x that fully loaded. Legacy TMS and ERP platforms added a second constraint on top: built to move a shipment or post a payment, not expose every rate and contract term to an outside check.

‍

Sampling wasn't chosen because 15-30% was the right amount of scrutiny. It was the largest percentage the labor budget, and the data access budget under it, could cover.

‍

What did that constraint actually cost, in a real case?

A Fortune 10 consumer technology company audited only 33% of its freight invoices, and $263M in accessorial charges passed through unvalidated every year.

Threshold-based systems couldn't handle the complexity of its shipments, multiple carriers, currencies, and specialized handling, across 5 million shipments a day and 150+ carriers in 40+ countries.

That gap wasn't a judgment call about which invoices mattered least. It was what the existing audit infrastructure could technically process. After moving to full coverage, that company went from 33% to 100% invoice coverage and recovered $3M+ annually through AI-powered freight costing accuracy, money that had been sitting in the unaudited 67% the entire time.

‍

What actually changed to make 100% coverage possible?

Checking an invoice stopped requiring a person to check it.

The labor-cost math that forced sampling assumed every additional invoice needed a human review cycle. An AI Team that resolves the exception itself, instead of routing it to a queue for a person to open, breaks that link between invoice volume and headcount.

That's an entirely different cost structure, the actual shift underneath the speed. Coverage stops being rationed once it doesn't scale with the size of your AP team.

‍

‍

What does full coverage actually recover that sampling misses?

Full-coverage AI-native auditing recovers 1.5 to 2.5% of total freight spend a year, money a sampled audit structurally cannot reach, since the invoices carrying that money were never opened in the first place.

That's a bigger gap than a marginal improvement over a well-run sample. It's the difference between checking a fraction of your freight spend and checking all of it.

The percentage a company chose to sample was never really a choice about risk tolerance. It was the size of the check the existing audit process could afford to write. Once that's no longer true, the sample doesn't need defending. It just needs replacing.

‍

{{brown-cta}}

‍

Frequently Asked Questions

Why do most freight audits only check 15-30% of invoices?

Historically, checking every invoice by hand meant paying the same labor cost on every single line, whether or not it contained an error. Sampling was the largest coverage a manual audit team's headcount budget could support, not a deliberate risk strategy.

What percentage of freight invoices actually contain errors?

It varies by company and carrier mix, but the risk isn't evenly distributed, which is exactly why a fixed percentage sample misses real money. A Fortune 10 company auditing only 33% of its invoices had $263M in accessorial charges passing through unvalidated every year.

Does full-coverage auditing cost more than a sample-based audit?

Not in the way headcount-based auditing did. Full coverage recovers 1.5-2.5% of total freight spend precisely because it doesn't require adding a person for every additional invoice checked.

Is sampling ever still the right approach?

For very low invoice volume and minimal complexity, a manual sample may still be workable. The constraint that made sampling necessary at scale, headcount scaling with invoice volume, is what full-coverage AI auditing removes.

^Sources

• Ardent Partners, State of ePayables
• ZipRecruiter, Accounts Payable Clerk Salary
• U.S. Small Business Administration
• Freehand, Invoice Audit Agent
• Freehand, Accounts Payable Outsourcing: Cost, Pros and Cons, and the AI Alternative
• Nitin Jayakrishnan, The Supply Chainer
• Freehand, Fortune 10 Consumer Tech Company Freight Audit Transformation