Termco
Termco

Nobody Disputes an Undercharge

19 November 2026

On the same Thursday afternoon, the same clerk keys two bad tickets.

Both are honest mistakes of the same species. Ticket 4418, Kenner Farms, 23.6 tons of road base, goes into the system as 26.3. A transposition, the kind of slip every keyboard produces at volume. Fourteen tickets later, ticket 4431, Pruitt Paving, 26.4 tons of 3/4 clean, goes in as 24.6. Same slip, opposite direction. At $11 a ton, Kenner is now overbilled $29.70 and Pruitt is underbilled $19.80.

If keying errors were coin flips, and they roughly are, this Thursday was a fair one. Follow each error for two weeks and watch the fairness disappear.

The error that comes home

Kenner's invoice lands on the first of the month. Their bookkeeper does what AP people everywhere do: she pulls the driver's copies from the job folder and matches them line by line. Ticket 4418 says 23.6 tons in the scale operator's handwriting. Your invoice says 26.3. She calls that afternoon.

Your office pulls the yellow copy, reads the same 23.6, and agrees on the spot. A credit memo goes out for $29.70, somebody writes "keying error" in the memo line, and the correction posts before the month closes. Add it up and two offices spent maybe twenty minutes of combined labor to fix a thirty-dollar mistake. Annoying, but the system worked. The error existed for eleven days and then it didn't.

The error that doesn't

Pruitt's AP clerk runs the exact same match on ticket 4431. Her delivery copy says 26.4 tons. Your invoice says 24.6. The invoice is lower than her number, and here the paths split, because there is no AP department on earth whose job is to raise your invoices.

Her job is to protect her company from paying too much. An invoice that comes in under the backup isn't a discrepancy to her, it's a Tuesday. Some clerks will circle it and call, and if you have a customer like that, hold onto them. Most approve the line and move on. Plenty of three-way-match software is configured to auto-approve any line at or below the matched quantity, so the rule itself waves the undercharge through.

No call comes. No document is created. The $19.80 doesn't show up as a dispute, a write-off, or a variance. It shows up as nothing, which is why you've never seen it.

Symmetric errors, one-way corrections

Here's the claim worth arguing about: your error rate is not the problem. The direction of your corrections is.

A clerk keying 180 tickets a day at 99 percent accuracy, which is a good clerk having a good day, produces roughly two bad tickets daily. Keystroke slips don't favor a direction, so over a month that's about twenty errors high and twenty low. The high ones get the Kenner treatment: matched, called, credited. Not all of them, and the small accounts that never match a line leave errors standing in both directions. But your bigger customers' AP departments are the most thorough auditors your invoices will ever meet, they work for free, and they check every line in one direction only.

The low ones get the Pruitt treatment.

So the errors that get corrected are overwhelmingly the ones that favored you, and the errors that survive are overwhelmingly the ones that didn't. After the dust settles, your billing doesn't carry random noise. It carries a bias, and every ounce of it points away from you.

flowchart TD
    T["Keying tray: 180 tickets"] --> K["Clerk keys the day's tickets"]
    K --> H["Error keys high: 26.3 for 23.6"]
    K --> L["Error keys low: 24.6 for 26.4"]
    H --> M1["Customer AP matches invoice to delivery copy"]
    L --> M2["Customer AP matches invoice to delivery copy"]
    M1 --> D["Dispute call"]
    D --> C["Credit memo: error corrected"]
    M2 --> S["Line approved, no call"]
    S --> G["Underbilling never surfaces"]

Your credit memo file has a twin

This asymmetry hands you a measuring stick, because the visible half of a symmetric distribution tells you the size of the invisible half.

Pull ninety days of credit memos and keep only the ones caused by keying: wrong quantity, wrong product code, transposed digits. Total them. If customer-caught keying errors cost you $1,400 in credits last quarter, then something near $1,400 in undercharges from the same keyboard left the building over the same quarter, and nobody called about those. That's $5,000 to $6,000 a year from one ordinary, competent clerk, before you count the combined labor each visible error burns on top.

Those numbers are illustrative. Yours might be smaller. The way to know is the second half of the test: sample fifty invoiced lines against the original ticket copies and count the misses in each direction yourself. If you find highs but no lows, your clerk has a talent worth studying. You'll find both.

The fix is upstream of the keyboard

You can't coach your way out of this, because the clerk isn't the flaw. Retyping is. Every number that moves from handwriting to keyboard rolls the dice again, and you've already seen which way the house edge runs.

This is the actual reason Termco digitizes tickets instead of replacing them. The AI reads the paper ticket your operation already produces, handwriting and all, so the quantity on the invoice is the quantity on the ticket, not a retyped copy of it. Tickets that don't add up get flagged instead of keyed past. No driver installs anything, the carbon books stay on the dash, and the scale window works the way it worked last week. If you want to see what's sitting in your own tray, send 10-20 sample tickets to hi@termco.ai and you'll have findings within 48 hours.

Either way, the credit memo file is worth an hour this week. Every memo in it is a receipt for an error a customer caught for you. The customers will keep finding the high ones. The low ones are yours to find.