Two Loads That Look Like One

Two belly dumps run a nine-mile loop all morning: pit, county road job, back. Ticket 41277 prints at 10:12 for 23.9 tons of base. Ticket 41283 prints at 10:41 for 23.6 tons. Same product, same job, same rate. At month end they land on the invoice as lines 118 and 119, one right above the other, $312 and $308. The contractor's project accountant reads down the column, stops at the near-twins, and does what she's paid to do. She marks the pair, pays one, and short-pays the other with a note: duplicate billing, credit ticket 41283.

Both loads ran. You know it, the drivers know it, and the base is compacted into the road. Now look at what the paper knows. The two tickets agree on product, job, rate, and weight within a few hundred pounds. The truck field reads T-7 on both, because the scale operator keyed the second weighing without changing the truck from the last one, the way hands do during a rush. The driver line is blank on one and a scrawl on the other. So the single fact that made these two loads instead of one, two different trucks 29 minutes apart, exists nowhere on the documents. A miskey in a field nobody prices from just erased the field you needed most.

The alibi fields

A ticket carries two kinds of information. The price fields, tons and product and rate, get read by everyone, because the total comes from them. Then there are the fields that don't move the total at all: time, truck number, driver, tare weight, signature. Call those the alibi fields. They don't say what the load cost. They say the load happened, this one distinctly from every other one.

Nobody proofs the alibi fields, precisely because they don't change the invoice. A scale operator who fat-fingers a rate gets caught by month end. One who leaves T-7 sitting in the truck field all morning gets caught by nobody, since every total still comes out right. The error waits. It only matters on the day a customer asks whether lines 118 and 119 are the same event, and on that day it's the entire case. T-7 tares at 14,220 pounds. The truck that actually hauled 41283 tares at 15,060. One glance at the tare column would settle the question, if the ticket carried the right truck.

Why the credit gets cut anyway

The disputed line is $308. Proving it means someone pulls the scale roll, finds both weighings, matches them to dispatch's morning board, and maybe calls the foreman who signed. Call it an hour of office time, spent arguing with a customer who pays you $40,000 a month, over a line they've already decided looks wrong. Most offices don't spend the hour. They cut the credit, note it as a billing adjustment, and move on. Cheaper to credit. Nobody wrote that policy down, and everybody applies it.

Here's what that buys. The customer's AP department now has a data point: duplicate flags against your invoices return credits on the first attempt, no fight. So the column gets read with a sharper pencil next month. And honest near-twins are everywhere in this work, because that's what steady production looks like. A loader that fills a truck every 22 minutes, two trucks on rotation, one product, one job: the operation manufactures look-alike pairs all day long. Each one is now a coin flip, and you fund both sides of it.

Some flags are right, which is the uncomfortable part. Real duplicates exist in your stack. The scale copy gets keyed at the window on Monday, the driver copy gets keyed Thursday by somebody clearing the turn-in pile, and serial 41283 goes out doubled. The customer catches those, and should. Plenty of owners shrug here: the customer catches it, we credit it, no harm done. The claim worth arguing about is that a true duplicate is the most expensive line you send, and not because of the doubled $308, which was never your money. It's expensive because it proves to the other side that flagging works. Every real duplicate you let out the door funds the harvest of your real pairs.

flowchart TD
  A["Ticket keyed into billing"] --> B{"Serial 41283 already in the system?"}
  B -- "nobody checks" --> C["Invoice doubles the load"]
  C --> D["AP catches it: credit memo issued"]
  D --> E["Next month every near-twin gets flagged"]
  E --> F["Honest pair, truck field miskeyed: $308 credited"]
  B -- "flagged at capture" --> G["Duplicate dies before the invoice"]

Give every load its alibi

The fix runs in two directions, and both belong to your side of the fence.

Kill the real duplicates before they travel. Tickets digitized as they're captured can be checked against everything already captured, so the second appearance of serial 41283 gets flagged as non-compliant that day, while the Thursday keyer still remembers the pile it came from. A duplicate that never reaches an invoice never teaches your customer anything.

Then make the pairs provable. The alibi fields stop being optional when every ticket gets checked against reference data instead of against nothing: a truck list with tares on it, the drivers, the jobs each truck is assigned to. A ticket that says T-7 but carries a 15,060-pound tare doesn't match the truck list, so it gets flagged the week it's captured, and the fix costs a radio call. Catch the same miskey in a short-pay four months later and it costs $308 and an argument you'll skip.

An invoice assembled only from tickets that cleared those checks goes out with truck, time, and tare intact. Line 119 can prove it isn't line 118. That's the slice Termco builds: tickets digitized at capture, non-compliant ones flagged, duplicates included, billing rules structured, reference data managed, and invoice totals assembled from nothing else. Whether you fight any given $308 flag stays a judgment call. It just stops being an hour of archaeology and becomes two fields read off a screen.

There's a blunt way to size this leak from your own books. Pull last quarter's credit memos and count the ones whose reason says duplicate. Then take five and answer, from your own paper, whether the credited load actually ran. Every yes is revenue you earned, hauled, invoiced, and then handed back because two tickets that told the truth couldn't prove they weren't each other.

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