The real cost of an expense error: modeling AP time with GBTA figures

Start with two published numbers

Most finance leaders know expense review is expensive, but few can put a number on it in a budget meeting. The Global Business Travel Association (GBTA) gives you a defensible starting point: an average of $58 to process one expense report, and $52 plus 18 minutes to correct a report that contains errors.

GBTA research also puts the share of reports containing errors or out-of-policy items at about 19%. Those three figures are enough to build a model your CFO can check line by line. The point is not precision to the dollar. The point is to replace a vague sense of pain with a number you can track month over month.

A worked model for a 900-person company

Take Calloway Instruments, a 900-person maker of lab measurement equipment with a large field sales and service team. Calloway's AP team receives 1,640 expense reports in a typical month. Applying the GBTA error rate of 19% gives 311.6 reports, which we round to 312 reports that need correction.

At $52 per correction, that is 312 x $52 = $16,224 a month. At 18 minutes per correction, it is 312 x 18 = 5,616 minutes, or 93.6 hours of correction work every month. Annualized, Calloway spends $194,688 and about 1,123 hours fixing reports that were wrong when they arrived.

Now add the baseline. Processing all 1,640 reports at $58 each costs $95,120 a month, so corrections add roughly 17% on top of the processing cost. The 93.6 hours is a little more than half of one full-time AP specialist working a 160-hour month, spent entirely on rework.

• Reports per month: 1,640
• Reports with errors at 19%: 311.6, rounded to 312
• Correction cost: 312 x $52 = $16,224 per month, $194,688 per year
• Correction time: 312 x 18 minutes = 93.6 hours per month
• Baseline processing: 1,640 x $58 = $95,120 per month

What the model leaves out

The $52 figure covers the cost of correcting a report you caught. It says nothing about the reports you did not catch. A mid-market analysis of T&E programs found that 6% to 10% of T&E spend falls outside policy, and much of that is reimbursed because nobody looked closely enough.

Suppose Calloway's monthly T&E spend is $611,480.37. At 6% to 10%, between $36,688.82 and $61,148.04 a month is out of policy. Over a year, that is roughly $440,000 to $734,000, which dwarfs the correction cost. Even if only a fraction is recoverable, the leakage line belongs in the same model.

There is also a fraud tail. The ACFE Report to the Nations finds that expense reimbursement schemes make up 14% of asset misappropriation cases, with a median loss of $40,000 and a median of 14 months before detection. A single undetected scheme can cost more than a quarter of correction work.

Where the 312 corrections come from

When we break down correction queues with finance teams, the same few causes show up again and again. Knowing your own mix matters because each cause has a different fix, and some fixes cost almost nothing.

At Calloway, a two-week tally of 157 corrected reports showed that missing or non-itemized receipts led the list, followed by personal items inside hotel folios, meals coded to the wrong account, and duplicate submissions of the same ride or dinner. A small number involved threshold splits that needed a second approver.

• Missing itemized receipts for meals over the $75 substantiation line
• Personal folio charges such as spa, movies and minibar alcohol
• Client meals coded as entertainment, or the reverse
• The same Uber ride submitted on the corporate card and as an out-of-pocket claim
• Purchases split across two transactions to stay under an approval limit

Levers that actually move the number

There are three levers in the model: the error rate, the cost per correction and the share of leakage you catch. Lowering the error rate means catching problems before a report reaches AP, ideally while the employee still remembers the dinner. Lowering the cost per correction means making the fix request so clear that the employee can resolve it in one step.

Run a scenario. If Calloway's error rate fell from 19% to 11%, corrections would drop to about 180 a month, or $9,360 and 54 hours. That frees $6,864 and nearly 40 hours a month before you count any recovered leakage. These are scenario numbers for planning, not a forecast, but they show where the return sits.

Running the model on your own data

You can build this in a spreadsheet in an afternoon. Pull last quarter's report counts, count how many were sent back or edited after submission, and apply the GBTA cost figures. Then sample 50 approved reports and audit them line by line to estimate leakage.

Phyto Nex AI customers often run the model before and after rollout. CloudScale Technologies cut audit time by 85%, and across customers 94% of employees correct a flagged item themselves once they receive a clear, clause-cited message that finance has approved. The model is the easiest way to see whether that holds for your team.

• Count monthly reports and reports returned or edited after submission
• Multiply returned reports by $52 and 18 minutes
• Sample 50 approved reports and audit every line item
• Estimate out-of-policy spend and compare it to the 6% to 10% range
• Repeat the model quarterly and track the trend, not the absolute figure