The Clerk Who Never Makes Mistakes

Tuesday, 8:15 in the morning. A ticket for 24.2 tons of screened topsoil lands in the keying tray with job number 2117 written on it, and your senior billing clerk, call her Rhonda, keys it to 2171 without slowing down. She's right. Job 2117 was a paving base job that wrapped up in July, and Kessler Grading has been pulling topsoil for 2171 all month. The driver's handwriting transposed the digits at the scale, the way handwriting does.

Now find that correction anywhere in your operation. There's no log entry, no flag, no note on the ticket. The scale house never learns it wrote the wrong job down. As far as any record shows, nothing happened at 8:15, and that's the point: in most billing offices, the entire error-catching apparatus is one person's memory, and the evidence of everything it catches is silence.

The fixes nobody logs

Watch a twenty-year clerk work a stack and count the interventions. She keys a load to Lambrecht Excavating instead of Lambert Farm Supply because she knows which of the two look-alike accounts runs quad-axles. She pulls a ticket out of the pile because she remembers keying that serial on Thursday, off the driver copy, and this is the scale copy coming through the turn-in pile four days late. She notices 40318 is missing because she keys in serial order and the run jumped from 40317 to 40319, so she radios the pit before lunch. She prices a gate-quoted load at $11.50 because she was standing there in April when the sales manager said it, and no sheet ever caught up.

None of that is data entry. It's reconciliation, performed at reading speed, against a reference file that exists in one head. She isn't your fastest keyer. She's your duplicate check, your job-number validation, your serial audit, and about half your rate sheet, and none of those functions appear on her job description or anywhere else.

Here's the uncomfortable arithmetic that follows. The miskeys are still happening upstream, at whatever rate the scale house and the drivers produce them. Your invoices just don't show it, because every one of those errors dies quietly at her desk. The office looks clean. The process isn't.

Two weeks in October

Rhonda's knee surgery is scheduled for the second Monday in October. The stack doesn't stop, so the front-desk hire keys it, carefully, exactly as written. That's the correct way for her to do it. Keying what the paper says is the honest move when you don't carry twenty years of context, and she doesn't.

So for ten working days, the tickets bill what they say instead of what they meant. The transposed job number bills a job that closed in July, and the contractor's project accountant rejects the line in November. The late scale copy goes out as a second invoice line, and the customer marks the pair as duplicate billing and short-pays both while they're at it. Serial 40318 never gets radioed about, so whatever it carried leaves no trace at all. The gate-quoted load prices off the sheet at $9.25, two dollars and a quarter light on every ton.

Notice which errors come back. The overbills return as short-pays and credit requests, six weeks out, with your customer's pencil marks on them. The underbills and the missing loads return never. Nobody audits an invoice hoping to owe you more.

Put loose numbers on it, yours to swap. If she quietly repairs four or five tickets a day and three of them carry real money at an average of $80 a line, her absence costs about $240 a day, call it $2,400 for the two weeks. Add one doubled invoice on top, the kind that teaches a good customer to read you with a sharper pencil. Then extend it, because the surgery is a drill for the retirement. She's 61.

flowchart TD
  A["Ticket lands in the keying tray"] --> B{"Who keys it?"}
  B -- "the veteran" --> C["Repaired from memory, keyed right"]
  C --> D["Clean invoice, no record any catch happened"]
  B -- "the fill-in" --> E["Keyed exactly as written"]
  E --> F["Overbills come back as short-pays"]
  E --> G["Underbills and missing loads never come back"]

The better she is, the worse your position

Here's the claim to argue with: a billing office with no visible error rate is in worse shape than one that shows a few mistakes a week, because the quiet office has converted its process failures into key-person risk and lost the ability to see either one.

Most owners will push back hard. Rhonda costs less than any system, never complains, and hasn't sent a bad invoice in a decade. All true. But a control that lives in one person runs only when that person is in the chair, and she isn't in the chair on weekends, at 6:40 on the Saturdays your pit opens early, or during the October you just priced. Her catches also protect the upstream sloppiness that creates them. The scale house has been transposing job numbers for years at the same steady rate, and it's never once heard about it, because absorbing the error is faster than reporting it. You can't fix a leak your best employee keeps invisible.

And when she does retire, you won't hire her back. The posting will say billing clerk, data entry, attention to detail. What walked out was a truck list, a customer alias table, a job registry, and a decade of quoted rates. No candidate has those.

Write down what she knows

A better fill-in won't fix this, and neither will a binder she never has time to write. The checks have to move out of her memory and into structure, so they run no matter whose hands are on the keyboard.

The things she knows are reference data: which trucks belong to which hauler, which accounts look alike and aren't, which job numbers are open and which closed in July. The rates she carries are billing rules with effective dates, including every gate quote that currently lives nowhere. The checks she runs by feel become checks that run at capture. Tickets get digitized the day they're written, so a serial that's already in the system gets flagged as a duplicate that day, a job number that matches nothing open gets flagged before it bills, and a rate no written rule supports gets flagged instead of keyed on faith. Invoice totals then get assembled only from tickets that cleared. That machinery is the part Termco builds: digitizing the tickets, flagging the non-compliant ones, structuring the billing rules, managing the reference data, assembling the totals. Rhonda stops being the single point of failure and becomes the person who curates what the checks compare against, which is the job she's actually been doing all along, unpaid and unwritten.

There's a one-week test that costs nothing. Ask her to stop fixing silently and start flagging: every ticket she'd have repaired goes in a tray with a note saying what was wrong with it. Friday, count the tray and price it. That pile is your real upstream error rate, the one your invoices have been hiding, and it's also a preview of every week after she's gone.

If you'd rather have a second reader on it, send 10-20 sample tickets to hi@termco.ai and we'll send back what we find within 48 hours.